HABILITÁCIÓS PÁLYÁZAT DR. VASSÁNYI ISTVÁN
Pannon Egyetem Műszaki Informatikai Kar
2017
Tartalomjegyzék
Habilitációs kérelem ..................................................................................................................... 1 Tudományos önéletrajz ................................................................................................................. 2 PhD oklevél másolata.................................................................................................................... 6 Habilitációs pontok számítása ....................................................................................................... 7 Tudományos tevékenység igazolása ............................................................................................ 19 MTMT-ből exportált publikációs lista hivatkozásokkal .................................................. 20 Tananyagformáló tevékenység igazolása ..................................................................................... 41 Nyilatkozat az oktatói tevékenységről............................................................................ 42 A kidolgozott tárgyak tematikái ..................................................................................... 43 Neptun rendszerből exportált kurzuslista ....................................................................... 55 Javaslat tantervi előadások témájára ............................................................................................ 59 Javaslat kollokviumi előadásra .................................................................................................... 60 Nyilatkozat habilitációról ............................................................................................................ 61 Nyilatkozat fogadókészségről...................................................................................................... 62 Nyilatkozat az Informatikai Tudományok Doktori Iskola követelményeinek teljesítéséről ........... 63 A 10 legfontosabbnak ítélt publikáció listája ............................................................................... 66 A 10 legfontosabbnak ítélt publikáció másolata ........................................................................... 67
1
Vassányi István tudományos önéletrajza Születési hely és idő: Állampolgárság: Családi állapot:
Veszprém, 1969.07.30. magyar nős
Jelenlegi munkahely és beosztás: Pannon Egyetem, Műszaki Informatikai Kar, Villamosmérnöki és Információs Rendszerek Tanszék, egyetemi docens Elérhetőség:
8200 Veszprém, Egyetem u. 10. Tel: 06-88-624000/6136, Fax: 06-88-624526 e-mail:
[email protected] , web: http://virt.uni-pannon.hu
Legmagasabb iskolai végzettség, szakképzettség: okleveles villamosmérnök, BME PhD, műszaki informatika tudomány, BME
(végzés éve) 1993 2000
Idegen nyelvtudás: felsőfokú “C” típusú állami nyelvvizsga angol nyelvből középfokú “B” típusú állami nyelvvizsga német nyelvből középfokú “C” típusú állami nyelvvizsga orosz nyelvből Angol-magyar szakfordító és tolmácsdiploma, BME Eddigi munkahelyek beosztással: 1993-96 PhD hallgató, BME Villamosmérnöki és Informatikai Kar 1996-97 tudományos segédmunkatárs, KFKI Mérés- és Számítástechnikai Kutató Intézet 1997egyetemi tanársegéd, adjunktus, majd egyetemi docens a Pannon Egyetem Műszaki Informatikai Kar Villamosmérnöki és Információs Rendszerek Tanszékén. Díjak, kitüntetések: 2015 Mestertanár aranyérem, OTDT 2015 Pro Educatione díj, PE MIK Oktatási tevékenység: 1997-2001 Integrált áramkörök (ea.) 1998-2000 Méréselmélet (gyak.) 1999Adatbáziskezelő rendszerek alkalmazása (ea. + gyak.) 1999Információ- és hírközléselmélet (ea.) 2002-2008 Információs rendszerek alapjai (ea. + gyak.) 2009Adatbáziskezelő rendszerek megvalósítása (ea. + gyak.) 2010Információs rendszerek biztonságtechnikája (ea.) 2011Digitális technika I-II. (ea.) 2011Egészségügyi adatbázisok (ea. + gyak.)
2
Sikeresen lezárt nagyobb kutatási témák: •
•
•
•
•
•
•
Telemedicina fókuszú kutatások orvosi matematikai és informatikai tudományterületeken (TÁMOP-4.2.2.A-2011/1/KONV-2012-0073 sz. projekt). Futamidő: 2013. febr.-2015. május. Partnerek: Szegedi Tudományegyetem (konzorcium-vezető), Pannon Egyetem, MTA Szegedi Biológiai Kutatóközpont. A projektben betöltött szerep: az Életmód-elemzés és -tanácsadás alprojekt szakmai vezetése és menedzselése CSI: Central Nervous System Imaging (szerz. szám: ENIAC_08-1-2011-0002). Futamido: 2009-2013. Partnerek: ST Microeletronics (koordinátor), Philips Advenced Technologies, Philips Healthcare, AustriaMicroSystems, Guger Technologies OEG, MAT-TECH BV, Pannon Egyetem, Debreceni Egyetem, MTA Atommagkutató Intézet, Austrian Institute of Technology, Stichting IMEC-NL, Politecnico di Torino, University od Bologna, Kempenhaeghe Clinics. A projektben betöltött szerep: a Pannon Egyetem munkájának témavezetése és menedzselése HealthTrack: Telenor Objects alapú M-Health alkalmazás fejlesztése (4800006223 sz. ipari K+F megbízás) Futamidő: 2011-2012. Partnerintézmények: Telenor Hungary Kft, Szegedi Tudományegyetem, Szoftvertechnológia Tsz. A projektben betöltött szerep: szakmai vezető. Számítógéppel segített, személyre szabott étrendtervezés (Magyar-Szlovén TéT projekt, OMFB00234/2010) Futamidő: 2010-2012. Partnerintézmény: Jozef Stefan Institute, Ljubljana. A projektben betöltött szerep: témavezető. HomeHealth: otthoni távmonitorozó és felügyeleti rendszer kifejlesztése (NTP-Jedlik projekt, AALAMSRK OM-00191/2008) Futamidő: 2008-2011. Koordinátor: GE Healthcare, partnerek: PE, Szegedi Tudományegyetem, Budapesti Műszaki Főiskola, Mednet 2000 Kft, Meditech Kft. A projektben betöltött szerep: a PE szakmai munkájának menedzselése, a tudásmodellezési csoport vezetése. NEUROWEB: Integration and sharing of information and knowledge in neurology and neurosciences (Sixth Framework Programme, Contract no.: 518513. Koordinátor: Istituto Nazionale Neurologico Carlo Besta, Milano, további partnerek: Univ. of Patras, Erasmus University, Rotterdam, OPNI-AOK, Univ. of Milano, ...) Futamidő: 2006-2008. A projektben betöltött szerep: a Pannon Egyetem munkájának szakmai vezetése. (17056-252) Életmód- és táplálkozás-tanácsadó szakértői rendszer (Innocsekk projekt, INNO-6-2007-0021 OMFB-01013/2007) Futamidő: 2007-2008. A projektben betöltött szerep: a Pannon Egyetem munkájának szakmai vezetése. (17056-252)
•
Életmód-tanácsadó szolgáltatások megvalósítása az origo portálon Futamidő: 2005-2009. A projektben betöltött szerep: a projekt szakmai vezetése.
•
Intelligens fiziológiai állapotmonitorozó és távfelügyeleti rendszer (GVOP-3.1.1.-2004-05-0196/3.0, Koordinátor: MFA) Futamidő: 2004-2007. A projektben betöltött szerep: közreműködő kutató
•
Költség-hatékony egészségmegőrzés és gyógyítás információtechnológiai módszerekkel (NKFP OM 2/052/2001 sz. projekt, projektvezető: Dr. Kozmann György). Futamidő: 2001-2004. A projektben betöltött szerep: az 1. részfeladat (Internet bázisú, rizikó- és életmód-elemző és tanácsadó rendszer) vezetése.
•
Új módszerek az egészségügyi információ-tárolásban és -megjelenítésben (OTKA F037416 kutatási program). Futamidő: 2002-2003. A projektben betöltött szerep: témavezető.
•
Intelligens adatelemző központ létrehozása (IKTA-5, 00142/2002 sz. projekt) Futamidő: 2003. jan. 1.-2005. júl. 31. A projektben betöltött szerep: témavezető.
•
Formális tervezési módszerek fejlesztése FPGA-alapú processzortömbökhöz (FPGA: FieldProgrammable Gate Array, a felhasználó által programozható kaputömb. OTKA T022115). Futamidő: 1997-99. A projektben betöltött szerep: témavezető.
3
•
•
•
Újrakonfigurálható számítógép-struktúrák (FKFP 0410/1997 sz. MKM kutatási program) Futamidő: 1998-99. Fővállalkozó: BME Méréstechnika és Információs Rendszerek Tsz. A projektben betöltött szerep: a Veszprémi Egyetemen folyó munka témavezetése. Communication Processor Design (COPRODES) CP 940453 sz. európai Copernicus együttműködés, melynek célja egy automatikus, idődiagram-specifikáció alapján FPGA-programot előállító CAD eszköz kifejlesztése. Partnerek: Passaui Egyetem, Rigai Egyetem, Prágai Műszaki Egyetem, ASICentrum kft (Prága), és a Budapesti Műszaki Egyetem Folyamatszabályozási Tanszéke. Futamidő: 1996-99. A projektben betöltött szerep: közreműködő kutató. Development of straightforward methodology, tools and component library for FPGA-based implementation of image processing primitives belonging to early vision. PMFB TéT D/116/95 sz. magyar-német kormányközi együttműködés, futamidő: 1995-97. A projektben betöltött szerep: közreműködő kutató.
Fontosabb kutatási eredmények: Finom granularitású tömbprocresszorok megvalósítása FPGA-n A felhasználó által konfigurálható kaputömbök (FPGA-k) lehetővé teszik tetszőleges digitális hardver architektúra alacsony költségű és flexibilis megvalósítását. Az FPGA-alapú tömbprocesszorok automatizált tervezésére a hagyományos particionálás-elhelyezés-összeköttetési (PPR) algoritmusok (mint pl. min-cut, szimulált hűtés, rip-up and reroute stb.) alkalmazásával és továbbfejlesztésével kidolgoztam egy olyan tervezési módszertant, mely speciális, tórusz alakú felületen dolgozik, ezért valamennyi szomszédsági és összeköttetési elvárást egyszerre tudja kielégíteni, és egyetlen processzorelem (cella) megtervezésével a nem közvetlen szomszédok közti összeköttetéseket is képes megvalósítani. A módszert egy PPR eszköz formájában megvalósítottam, és bináris morfológiai célhardver tervezésére alkalmaztam. A kutatás időszaka: 1993-2000 Két szintű adatmodellek alkalmazása az egészségügyi adatcserében Az archetípusokra épülő kétszintű adatmodellek lehetővé tehetik a szakértői tudás és az informatikai referencia-modell szétválasztását az egészségügyben használt komplex adatstruktúrákat modellezése során, azonban ezek megvalósítására a kutatás idején nem volt még példa. Megmutattam és egy hazai egészségipari kommunikációs szabványon demonstráltam, hogy az XML/XSD technológia megfelelő módon alkalmazva elegendő leíró erővel rendelkezik mind a referencia információs modell, mind a ráépülő archetípusok megvalósítására, és szakértőként részt vettem a kétszintű modellt használó MSZ 22800 egészségügyi üzenő szabvány kidolgozásában. Ez a szabvány a jelenleg Magyarországon bevezetés alatt álló egészségügyi intézményközi kommunikációs rendszer egyik alappillére. A kutatás időszaka: 2000-2004 Az eredményt ismertető publikáció a 10 legfontosabbnak ítélt publikáció közül: [1] Általános célú adatbányászati szolgáltatás tervezése és megvalósítása Az adatelemzés, adatbányászat általában speciális szaktudást igényel és speciális szoftver eszközök és módszerek alkalmazását követeli meg a nem informatikus szakterületű szakértőktől. Témavezetőként kollégáimmal közösen olyan webes adatbányászati szolgáltatást hoztam létre, mely a szakértők elől elfedi a számára lényegtelen részleteket, és az adatok feltöltése után tipikus adatbányászati műveletsorozatok végrehajtását támogatja. A különféle szakterületek számára tervezési mintákat hoztunk létre az adatbányászati folyamatra. Az elgondolás használhatóságát szeizmológiai, régészeti és egészségügyi szakterületen végzett elemzésekkel demonstráltuk. A kutatás időszaka: 2003-2005
4
Neurológiai ontológiára alapozott adatföderációs megoldás tervezése és megvalósítása A neurológiai klinikákon létrejövő adatvagyon kutatási célokra való felhasználhatóságát a kis esetszámok korlátozzák, a klinikai adatbázisok föderációja ezért minden klinika érdeke. A Neuroweb nemzetközi együttműködésben kidolgozott neurológiai információ-megosztó és –feldolgozó keretrendszerben megvalósítottam a központi szakértői tudásmodell (ontológia) összekapcsolását a relációs sémába konvertált klinikai adatbázisokkal. A kidolgozott megoldás újdonsága, hogy automatizálja a különböző szintű, általánosított neurológiai fogalmak (ún. fenotípusok) felbontását a klinikai adatbázisok által támogatott adatelemekre, ezáltal támogatja a lekérdezések automatikus általánosítását a szakterületi logika szerint. A kutatás időszaka: 2006-2009 Az eredményt ismertető publikáció a 10 legfontosabbnak ítélt publikáció közül: [5] Egészségügyi ellátási eseménysorok innovatív elemzésére alkalmas informatikai módszertan kidolgozása és megvalósítása Bár a központi társadalombiztosítási adattárház (ÁEEK adattár) minden államilag finanszírozott esetről tartalmaz évtizedekre visszamenő, értékes adatokat, elemzési célokra való felhasználása csak a közelmúltban kezdődött el. Dr. Kósa Istvánnal közösen olyan módszertant dolgoztunk ki, mely egy adott szakterületen az ÁEEK adattár felhasználásával lehetővé teszi az adatok szakértői szabályok szerinti tisztítását, orvosi szempontból értelmes eseménysorok azonosítását, ezáltal az esetek tipizálását, majd az esetek területi alapon de facto ellátó klinikákhoz rendelése révén az egyes klinikák illetve ellátási régiók ellátási gyakorlatának statisztikai alapú jellemzését. Bár az alkalmazott statisztikai elemzési módszerek a szokásosak, de az adatok előkészítése és az ellátó rendszer ilyen jellegű jellemzése (részben a felhasznált adattár teljessége miatt is) nemzetközi viszonylatban is újszerű eredmény. Az általam megvalósított, relációs adatbázis alapú informatikai modellre jellemző az egész adatfeldolgozási folyamat paraméterezhetősége és a nagy adattömegek hatékony kezelése. A kutatás időszaka: 2010Az eredményt ismertető publikációk a 10 legfontosabbnak ítélt publikáció közül: [6,7] Életmód-támogató szakértői rendszer modellezése A megfelelő életmód kialakítása sok krónikus betegség megelőzésének illetve sikeres terápiájának előfeltétele. A helyes életmód elsajátításában különösen az öregedő nyugati társadalmakban egyre nagyobb szerepet vállalhat az informatika, feltéve, hogy a szakértői tudás megfelelően modellezhető. A területen végzett több mint tíz éves kutatás-fejlesztés során témavezetőként olyan rendszer kidolgozását irányítottam, amely elsősorban táplálkozási naplózást és a naplók szakértői rendszer általi értékelését támogatja. A rendszer a Lavinia életmód-tükör mobil alkalmazás formájában a Pannon Egyetem által licenszált piaci hasznosításra is került 2016-ban. A szolgáltatás fő célcsoportja az I. és II. típusú cukorbetegek illetve fogyni/hízni vágyók. Az általam kidolgozott informatikai modell elemei a tápanyagok, élelmiszerek, gyógyszerek, fizikai aktivitások, egymásba ágyazható receptek, betegségek illetve betegtípusok, mint fő entitások, és az ezekre épített hierarchikus taxonómiák/ontológiák illetve az egyes osztályok kapcsolataira kimondott szakértői szabályok. A kutatás külön ágát képezi az életmód (táplálkozás, gyógyszerezés, fizikai aktivitás, stressz), mint bemeneti adatok alapján a vércukorszint várható alakulásának rövid távú előrejelzése személyre szabott matematikai modellek felhasználásával. Ez a szolgáltatás jelentősen könnyítheti és biztonságosabbá teheti a cukorbetegséggel élők mindennapi életét. A másik specializált kutatási ág az adott elvárásokat kielégítő, személyre szabott étrendgenerálás több szintű, több szempontú genetikus algoritmusok segítségével, melyből PhD disszertáció is született. A szakértői rendszert dietetikus és orvos szakértők közreműködésével töltöttük fel, és öt klinikai vizsgálatban teszteltük cukorbetegek körében, a Lavinia alkalmazással. A vizsgálatok igazolták a tanácsadó rendszer használatához köthető egészség-nyereséget és az általános felhasználói elégedettséget. A kutatás időszaka: 2002Az eredményt ismertető publikációk a 10 legfontosabbnak ítélt publikáció közül: [2,3,4,8,9,10]
5
6
AZ OKTATÓI, KUTATÓI ÉS TUDOMÁNYOS KÖZÉLETI TEVÉKENYSÉG ÉRTÉKELÉSÉNEK SZEMPONTJAI /A pályázó önértékelése a PhD vagy kandidátusi fokozat megszerzését követően, de legfeljebb az utolsó öt évben végzett tevékenység alapján. Az egyes pontokat tételes, beazonosítható módon kell felsorolni./ PhD fokozat megszerzésének éve: 2000 A pontszámításhoz felhasznált adatok a 2012-től 2016-ig terjedő időszakra vonatkoznak. Összes pontszám: 3220.77 (min. 1320)
I. OKTATÁSI TEVÉKENYSÉG Összesítés: I.
OKTATÁSI TEVÉKENYSÉG
pontszám 2012 2013 2014 2015 2016 összesen
1.
Gyakorlat tartása (1 óra)
1
68
68
124
68
68
396
2.
Előadás tartása (1 óra)
2
170
148
232
186
170
1812
3.
Doktorképzési kurzus (1 óra)
3
4.
Megvédett szakdolgozat témavezetése magyar nyelven
5
7
5
2
5
13
160
5.
TDK konzulensi tevékenység 2
4
120
országos
20
egyetemi(kari)
10
6.
Megvédett PhD (asp.) témavezetése magyar nyelven
30
7.
Tantárgyi program készítése (programonként)
20
1
Az oktatók minimális oktatói teljesítménye: 1000 A főállású kutatók minimális teljesítménye: 400
20 2508
Részletezés: 1. Gyakorlat tartása (bővebben lásd a Neptun kurzuslistáját) •
VEMKSA5144A Adatbázis-kezelő rendszerek alkalmazása, 14*2 óra gyakorlat félévenként, 5 félév: 140 pont
•
VEMIVI5154A Adatbázis-kezelő rendszerek megvalósítása, 14*2 óra gyakorlat angol nyelven 2014 tavaszán: 56 pont
•
VEMKSA5144A Adatbázis-kezelő rendszerek alkalmazása, levelező kurzus, 12 óra gyakorlat félévenként, 5 félév: 60 pont
7
•
VEMIVI5154A Adatbázis-kezelő rendszerek megvalósítása, 14*2 óra gyakorlat félévenként, 2012, 2014, 2016 tavaszi félévében: 112 pont
•
VEMIVIM554E Egészségügyi adatbázisok, 14*2 óra gyakorlat 2015 tavaszi félévében: 28 pont
Gyakorlat összesen: 396 pont 2. Előadás tartása (bővebben lásd a Neptun kurzuslistáját) •
VEMIIR3112I Információ- és hírközléselmélet, nappali kurzus, 14*2 óra előadás félévenként, 5 félév: 280 pont
•
VEMIIR3112I Információ- és hírközléselmélet, levelező kurzus, 16 óra előadás félévenként, 5 félév, 2015 őszi félévében angol nyelven: 192 pont
•
VEMIVI1112D Digitális technika I, 14*2 óra előadás félévenként, 5 félév: 280 pont
•
VEMIVI2112D Digitális technika II., 14*2 óra előadás félévenként, 5 félév: 280 pont
•
VEMKSA5144A Adatbázis-kezelő rendszerek alkalmazása, 14*2 óra előadás félévenként, 5 félév: 280 pont
•
VEMIVI5154A Adatbázis-kezelő rendszerek megvalósítása, 14*2 óra előadás angol nyelven 2014 tavaszán: 112 pont
•
VEMKSA5144A Adatbázis-kezelő rendszerek alkalmazása, levelező kurzus, 12 óra előadás félévenként, 5 félév: 120 pont
•
VEMISAM144B Információs rendszerek biztonságtechnikája, 8 óra előadás félévenként, 5 félév: 80 pont
•
VEMIVI5154A Adatbázis-kezelő rendszerek megvalósítása, 14*2 óra előadás félévenként, 2012, 2014, 2016 tavaszi félévében: 168 pont
•
VEMIVIM554E Egészségügyi adatbázisok, 14*2 óra előadás 2015 tavaszi félévében: 56 pont
Előadás összesen: 1812 pont 4. Megvédett szakdolgozat témavezetése magyar nyelven •
2012: Arnhoffer Emőke, Kárpáti János, Papp György, Töltési Dávid, Grinácz Tamás, Béres Mónika, Cseh Tamás (BSc): 35 pont
•
2013: Elek Csilla, Karikó Ivett, Szabó István, Tompos Ádám (BSc), Apostol Gergely (BSc): 25 pont
•
2014: Bódis Noémi, Koppány Attila: 10 pont
•
2015: Tompos Ádám (MSc), Apostol Gergely (MSc), Nagy Gergő, Varga Szabolcs, Gyuk Péter: 25 pont
•
2016 tavasz: Papp Boglárka, Szakács István, Baracska Péter, Cseh Tamás (MSc), Dömötör Róbert, Lőrincz Tamás, Mesterházi Sándor, Szabó Zénó: 40 pont
•
2016 ősz: Deés Lea, Dobos Ádám, Lénárt Bálint, Szabó Tamás, Tóth Kitti: 25 pont
Szakdolgozat összesen: 160 pont
8
5. TDK konzulensi tevékenység pályamunkánként, az OTDK-ra továbbjutott ITDK dolgozatokat csak egyszer számítva (bővebben lásd a kari TDK honlapot https://tdk.mik.uni-pannon.hu/index.php/ ): •
Egyetemi TDK konzultálás: Béres Mónika (2013), Gyuk Péter, Lőrincz Tamás (2014), Apostol Gergely, Tompos Ádám (2015), Mesterházi Sándor (2016), Gyuk Péter (2016, angol nyelven): 60 pont
•
Országos TDK konzultálás: Béres Mónika: (2015), Gyuk Péter, Lőrincz Tamás (2015), Gyuk Péter (2016-2017, angol nyelven), Mesterházi Sándor (2016-2017), Apostol Gergely, Tompos Ádám: (2016-2017): 120 pont
TDK összesen: 120 pont 7. Tantárgyi program készítése (bővebben lásd a Tananyagformáló tevékenység fejezetet) •
VEMIVIM554E Egészségügyi adatbázisok, 2015 tavaszi félévében: 20 pont
Tantárgyi program összesen:20 pont
II. PUBLIKÁCIÓS TEVÉKENYSÉG Összesítés: II PUBLIKÁCIÓS TEVÉKENYSÉG 1. Kongresszusi előadás és poszterkivonatok 1.1. Magyar nyelven hazai rendezvényeken 1.2 Idegen nyelvű hazai rendezvényeken 1.3. Nemzetközi hazai rendezvényeken 1.4. Külföldi rendezvényeken 1.5. Világ-konferencián vagy kongresszuson 2. Közlemények 2.1. Magyar nyelvű, ismeretterjesztő (nem hírlapban) 2.2. Magyar nyelvű, lektorált szakfolyóiratban 2.3. Magyar nyelvű, kongresszusi kiadványban 2.4. Idegen nyelvű, lektorált hazai folyóiratban 2.5. Idegen nyelvű, kongresszusi kiadványban 2.6. Idegen nyelvű, lektorált külföldi szakfolyóiratban 2.7. Nemzetközi tematikus bibliográfia (idegen nyelven) 2.8. Az utóbbi 5 év publikációinak összesített impakt faktora
pontszám 2012 2013 2014 2015 2016 Összesen
1
0
2
0
3 3 5
0 6 0
2
2
0
4
1
4
3
7
5
3
8
7
3
56
1
64
2
8
3
12
3
5
3
2
16 5
2
144
1
72
18
0
IF=pont
9
5.5
0.06
1.21
6.77
3. 3.1. 3.2. 3.3. 3.4. 4. 4.1. 4.2. 4.3. 4.4. 5. 5.1. 5.2. 5.3.
Tankönyv, egyetemi jegyzet Egyetemi jegyzet fejezete 8/ív Önálló egyetemi jegyzet 12/ív Egyetemi tankönyv fejezete 8/ív Önálló egyetemi tankönyv 12/ív Szakkönyv, vagy szakkönyv fejezete Magyar nyelvű könyv vagy 8/ív fejezete (ismeretterjesztő) Magyar nyelvű szakkönyv vagy 12/ív fejezete Idegen nyelvű szakkönyv vagy 16/ív fejezet hazai kiadónál Idegen nyelvű szakkönyv vagy 20/ív fejezet külföldi kiadónál Tudományos közleményekre hivatkozás Nemzetközi mérvadó kézikönyv 5 Idegen nyelvű lektorált folyóirat 2 Magyar nyelvű lektorált folyóirat 1 A főállású oktatók minimális publikációs teljesítménye: A főállású kutatók minimális publikációs teljesítménye:
0 0 0 0
0 0 0 0
5 17 170 425
25 34 0 423.77
A publikációk részletezése (bővebben lásd a Tudományos tevékenység igazolása c. fejezetet): CIKK CÍME ÉV CIKK TÍPUSA SCI IF MenuGene: A Comprehensive Expert System for 2012 ang_ny_konf_cikk Dietary and Lifestyle Counseling and Tracking
PONT 8
GPU based parallel genetic algorithm library Personalized Nutrition Counseling Expert System Regional differences in the utilisation of coronary angiography as initial investigation for the evaluation of patients with suspected coronary artery disease
2012 2012 2012
ang_ny_konf_cikk ang_ny_konf_cikk külf_abstract
8 8 3
The effect of waiting times on the patient pathways in the evaluation of patients with suspected coronary artery disease
2012
külf_abstract
3
Egészségügyi adatvagyon hasznosítása a stabil coronaria betegek ellátásának elemzésére
2012
m_ny_konf_f_cikk
4
Formalizing harmony rules for nutrition counseling GAME: GPU accelerated multipurpose evolutionary algorithm library
2013 2013
ang_ny_f_cikk ang_ny_f_cikk
12 12
10
Regional differences in the utilisation of coronary angiography as initial investigation for the evaluation of patients with suspected coronary artery disease
2013
ang_ny_f_cikk
Challenges and limits in personalized dietary logging and analysis
2013
ang_ny_konf_cikk
8
Dietary Logging and Analysis for Tele-Care Using Harmony Rules
2013
ang_ny_konf_cikk
8
SOLO: An EEG Processing Software Framework for Localising Epileptogenic Zones
2013
ang_ny_konf_cikk
8
Lokális agykérgi aktivitás mérése Laplace-típusú EEG térképezéssel: A felbontás vizsgálata modellezéssel
2013
m_ny_konf_f_cikk
4
Étel adatbázisok tartalmi eltéréseinek hatása a diéta naplózás pontosságára
2013
m_ny_konf_f_cikk
4
Személyre szabott táplálkozás-tanácsadó rendszer harmónia szabályokkal
2013
m_ny_konf_f_cikk
4
Táplálkozási adatbázisok átjárhatósága A regionális vizsgálati frekvenciák és a viszgálatra kerülok halálozási mutatójának összefüggése koszorúérbetegség gyanújával kivizsgálásra kerülo betegekben
2013 2013
m_ny_konf_f_cikk m_ny_konf_f_cikk
4 4
Kardiológiai rehabilitációs kezelésben részesülo illetve ilyen kezelésre potenciálisan jelölt betegek gyógyszerfogyasztásának összehasonlítása
2013
m_ny_konf_f_cikk
4
Életmód-változtatást támogató mobil informatikai alkalmazások
2013
m_ny_konf_f_cikk
4
Epileptikus gócok EEG alapú lokalizálását támogató szoftver környezet
2013
m_ny_konf_f_cikk
4
The impact of geographical distances to coronary angiography laboratories on the patient evaluation pathways in patients with suspected coronary artery disease. Results from a population-based study in Hungary
2014
ang_ny_f_cikk
Personalized Dietary Counseling System Using Harmony Rules in Tele-Care
2014
ang_ny_f_cikk
12
The Effect of the Waiting Times on the Patient Pathways for Patients with Suspected Coronary Artery Disease
2014
ang_ny_konf_cikk
8
Combined Model for Diabetes Lifestyle Support Preclinical tests of an android based dietary logging application
2014 2014
ang_ny_konf_cikk ang_ny_konf_cikk
8 8
11
5.5
0.06
12
12
Lifestyle Log Based Blood Glucose Level Prediction for Outpatient Care
2014
ang_ny_konf_cikk
8
A fast, android based dietary logging application to support the life style change of cardio-metabolic patients
2014
ang_ny_konf_cikk
8
Telemedical Heart Rate Measurements for Lifestyle Counselling
2014
ang_ny_mo_f_cikk
8
Reliability of telemedical Heart Rate meters Képfeldolgozás folyamata az EEG neuroesztétikai vizsgálatában
2014 2014
ang_ny_mo_f_cikk m_ny_konf_f_cikk
8 4
Rehabilitációra érdemes és ténylegesen rehabilitációra kerülo betegellátási utak elemzése
2014
m_ny_konf_f_cikk
4
Felhasználói tapasztalatok a Lavinia mobil életmódtükör alkalmazással
2014
m_ny_konf_f_cikk
4
Életmód-naplózás pontosságának elemzése kardiológiai rehabilitációra kerülo betegek körében
2014
m_ny_konf_f_cikk
4
Okostelefon alapú diéta-naplózó rendszer diabeteses betegek számára
2014
m_ny_konf_f_cikk
4
Kamrai szívizom-repolarizáció heterogenitás modellezéses vizsgálata
2014
m_ny_konf_f_cikk
4
A szívkatéteres laboratóriumoktól mért földrajzi távolság hatása az iszkémiás szívbetegség gyanújával ellátásra került betegek ellátási útjára
2014
m_ny_konf_f_cikk
4
Diabétesz életmód-támogatás vércukorszintelőrejelzéssel
2014
m_ny_konf_f_cikk
4
Kamrai szívizom-repolarizáció heterogenitás vizsgálat 2014 bioelektromos képalkotóval
m_ny_konf_f_cikk
4
Távmonitorozásra is alkalmas pitvari fibrilláció detektálási módszer
2014
m_ny_konf_f_cikk
4
Validation of a low cost telemedical stress monitoring system
2015
ang_ny_konf_cikk
8
Automatic stress detection using simple telemedical heart rate meters
2015
ang_ny_konf_cikk
8
Effectiveness of mobile personal dietary logging Diabetes Lifestyle Support with Improved Glycemia Prediction Algorithm
2015 2015
ang_ny_konf_cikk ang_ny_konf_cikk
8 8
Blood Glucose Level Prediction for Mobile Lifestyle Counseling
2015
ang_ny_konf_cikk
8
Flexibilis, eseményvezérelt keretrendszer mobil telemedicinális alkalmazásokhoz
2015
m_ny_konf_f_cikk
4
12
Glikémiás hatást befolyásoló életmódbeli, étrendi tényezok vizsgálata cukorbetegek vércukor szintjére
2015
m_ny_konf_f_cikk
4
Hosszú hatású inzulin kezelése vércukorszintelőrejelző modellben
2015
m_ny_konf_f_cikk
4
Vércukor-előrejelző modell klinikai validációja Étrendértékelés és -tervezés mesterséges intelligencia segítségével
2015 2015
m_ny_konf_f_cikk m_ny_konf_f_cikk
4 4
Fiziológiai paraméterek változása életmód támogató informatikai rendszer használata során
2015
m_ny_konf_f_cikk
4
Stabil anginás betegutak klaszterelemzése Stress Detection Using Low Cost Heart Rate Sensors Characterizing blood glucose response to specific meals in pre-diabetes: a small scale study
2015 2016 2016
m_ny_konf_f_cikk ang_ny_f_cikk ang_ny_konf_cikk
4 12 8
Changes in the spatial distribution of dominant IHD care providers over a 10 year period in Hungary
2016
ang_ny_konf_cikk
8
Táplálkozási trendek, szakmai ajánlások. Mi a jövő útja?: Életmód-támogató mobil alkalmazás szerepe a dietoterápiában
2016
m_ny_konf_f_cikk
4
Diabetesesek dietoterápiájának és önmenedzselésének támogatása mobilapplikációk használatával
2016
m_ny_konf_f_cikk
A betegek anonimitásának biztosítása a földrajzi elhelyezkedésre kiterjedő egészségügyi adatelemzések során
2016
m_ny_konf_f_cikk
4
mHealth szolgáltatás felhasználói igényének felmérése
2016
m_ny_konf_f_cikk
4
0.92
0.29
4
A hivatkozások részletezése (bővebben lásd a Tudományos tevékenység igazolása c. fejezetet):: HIVATKOZÁSOK 2012-16 KÖZÖTT CIKK CÍME
ÉV
SCI IF
Implementing processor arrays on FPGAs
1998
0.44
An Evolutionary Divide and Conquer Method for Long-Term Dietary Menu Planning
2005
0.97
Application of artificial intelligence for weekly dietary menu planning
2007
13
kézikönyvben
angol ny. f. cikkben
PONT
1
2
2
3
16
1
1
7
An ontological modeling approach to cerebrovascular disease studies: the NEUROWEB case.
2010
Personalized Nutrition Counseling Expert System
1.94
2
8
26
2012
1
2
GAME: GPU accelerated multipurpose evolutionary algorithm library
2013
2
4
Regional differences in the utilisation of coronary angiography as initial investigation for the evaluation of patients with suspected coronary artery disease
2013
1
2
5.5
A publikációk és hivatkozások további részletezését lásd az MTMT-ből exportált publikációs listán.
III. EGYÉB TEVÉKENYSÉG Összesítés: III.
Egyéb tevékenység
1.
Szakfordítói tevékenység
2.
Szerkesztői tevékenység
pontszám 2012 2013 2014 2015 2016 Összesen
2.1. Hazai konferencia előadáskivonat-kötet
3/5 ív
0
2.2. Nemzetközi konferencia előadáskivonat-kötet
6/5 ív
0
2.3. Hazai magyar nyelvű folyóirat/évfolyam
9/5 ív
0
2.4. Hazai idegen nyelvű folyóirat/évfolyam
12/5 ív
0
2.5. Konferenciaelőadás-kötet v. cikkgyűjt. (magyar)
9/5 ív
12
13
45
2.6. Konferenciaelőadás-kötet v. cikkgyűjt. (idegen) hazai megjelenéssel
12/5 ív
0
külföldi megjelenéssel
18/5 ív
0
18/5 ív
0
2.7. Külföldi folyóirat 3.
Tudományos közéleti, ill. konferencia szervezői tevékenység
3.1. Konferencia szekcióelnök
14
hazai
4
2
1
2
20
külföldi/hazai nemzetközi
6
0
világ
8
0
hazai tudományos rendezvény szervezése
6
0
nemzetközi tudományos rendezvény szervezése
12
1
hazai kongresszus szervezése
12
1
nemzetközi kongresszus szervezése
20
0
szimp. Szerv. Biz. Tag
4
0
szimp. Szerv. Biz. Elnök
8
24
szimp. Szerv. Biz. Titkár
6
0
konferencia Szerv. Biz. Tag
6
konferencia Szerv. Biz. Elnök
12
0
konferencia Szerv. Biz. Titkár
8
0
kongresszus Szerv. Biz. Tag
10
0
kongresszus Szerv. Biz. Elnök
18
0
kongresszus Szerv. Biz. Titkár
12
0
főszerkesztő
10
0
felelősszerkesztő
5
0
tag
3
0
4.1. Tudományos bizottsági tagság
1/év
5
4.2. Tudományos társaságban tisztségviselő
3/év
0
4.3. Tudományos bizottságban tisztségviselő
3/év
3.2. Konferenciaszervezés
12 1
24
3.3. Tudományos rendezvények szervezése
1
1
12
3.4. Tudományos folyóirat szervezőbizottságában
4.
Részvétel a tudományos közéletben
1
15
2
1
1
1
18
4.4. Tudományos társaság tiszteletbeli tagja
5/év
0
4.5. Tudománypolitikai bizottság tagja
1/év
0
4.6. Tudománypolitikai bizottság tisztségviselője
3/év
0
elnök
10
0
titkár
5
0
tag
3
0
opponens
10
0
elnök
5
0
titkár
2
0
tag
1
0
5.
Tudományos bírálói tevékenység
5.1. PhD értekezés műhelyvitán vagy nyilvános védésen bizottsági
5.3. PhD szigorlati bizottsági
6.
Lektorálás
6.1. Pályázatok
4/pályázat
6.2. Kézirat (tud. közl.)
4/ív
6.3. Szakkönyv
3/ív
0
7.1. Szabadalom
50
0
7.2. Know-how, szoftver
30
7.3. Államilag minősített fajta
50
7.
4
1
8
2
60
Tudományos eredmények gyakorlati hasznosulása
1
30 0
Főállású oktatók oktatói és publikációs tevékenységén túli munkája: minimum 150 Főállású kutatók oktatói és publikációs tevékenységén túli munkája: minimum 200
221
FŐÁLLÁSÚ OKTATÓK MINIMÁLIS ÖSSZES TELJESÍTMÉNYE: FŐÁLLÁSÚ KUTATÓK MINIMÁLIS ÖSSZES TELJESÍTMÉNYE:
3152.77
Részletezés: III/2. Szerkesztői tevékenység
16
1320 1025
2.5. Konferenciaelőadás-kötet v. cikkgyűjtemény (magyar) vagy sorozat monográfia, 9/5 ív Neumann Kollokvium, 2015, bővebben lásd http://neumann-kollokvium.njszt.hu/archive/2015/ , 12 ív: 21.6 pont • Neumann Kollokvium, 2013, bővebben lásd http://neumann-kollokvium.njszt.hu/archive/2013/ , 13 ív: 23.4 pont III/2 összesen: 45 pont •
III/3. Tudományos közéleti, ill. konferencia szervezői tevékenység 3.1. Konferencia (levezető) szekcióelnök - hazai • • • • •
Neumann Kollokvium, 2015, Adatvagyon hasznosítása / orvosi statisztika szekció, http://neumann-kollokvium.njszt.hu/archive/2015/ : 4 pont Neumann Kollokvium, 2015, Képfeldolgozás szekció: 4 pont Neumann Kollokvium, 2014, Telemedicína szekció, http://neumannkollokvium.njszt.hu/archive/2014/ : 4 pont Neumann Kollokvium, 2013, Távdiagnosztikai/terápiás szolgáltatások kialakításának szoftvertechnológiája szekció, http://neumann-kollokvium.njszt.hu/archive/2013/ : 4 pont IME Infokommunikációs Konferencia, 2013, Kutatás-fejlesztés szekció, http://www.imeonline.hu/konferenciak-idorendben.php : 4 pont
Összesen: 20 pont 3.2. Konferenciaszervezés - hazai tudományos rendezvény szervezése - nemzetközi tudományos rendezvény szervezése •
Hungary-Italy Bilateral Session, 2013. nov. 23., Veszprém, bővebben lásd http://neumannkollokvium.njszt.hu/archive/2013/ithu.html : 12 pont
- hazai kongresszus szervezése • Neumann Kollokvium, 2015, Veszprém: 12 pont • Neumann Kollokvium, 2013, Veszprém: 12 pont Összesen: 36 pont 3.3. Tudományos rendezvények szervezése - konferencia szerv. biz. tag • •
Neumann Kollokvium, 2016, Szeged http://neumann-kollokvium.njszt.hu/ : 6 pont Neumann Kollokvium, 2014, Szeged http://neumann-kollokvium.njszt.hu/archive/2014/ : 6 pont
Összesen: 12 pont III/3 összesen: 68 pont
17
III/4. Részvétel a tudományos közéletben 4.3. Tudományos bizottságban tisztségviselő • •
VEAB Egészségügyi Informatikai Munkabizottság, titkár, 2013- : 12 pont NJSzT Orvos-biológiai Szakosztály, vezetőségi tag: 2012-2013: 6 pont
III/4 összesen: 18 pont III/6. Lektorálás 6.2. Kézirat (tud. közl.) 4/ív • • • •
2012: 2 ív (angol ny.): 16 pont 2013: 1 ív: 4 pont 2015: 4 ív (angol ny.): 32 pont 2016: 1 ív (angol ny.): 8 pont
III/6. összesen: 60 pont III/7. Tudományos eredmények gyakorlati hasznosulása 7.2. Know how, szoftver •
Lavinia életmód-tükör. Kutatás, fejlesztés, hasznosítás, piacra vitel: 30 pont
Dokumentáció: http://lavinia.hu/ III/7 összesen:30 pont
18
A KÉRELMEZŐ TARTÓS TUDOMÁNYOS SZAKMAI, ALKOTÓ TEVÉKENYSÉGÉT ÉS SZAKIRODALMI MŰKÖDÉSÉT BIZONYÍTÓ DOKUMENTUMOK
19
Vassányi István (Informatika) 2017 1. REBAZ A H Karim , VASSÁNYI István , KÓSA István Blood glucose response characterization for outpatient pre-diabetes care IME: INTERDISZCIPLINÁRIS MAGYAR EGÉSZSÉGÜGY / INFORMATIKA ÉS MENEDZSMENT AZ EGÉSZSÉGÜGYBEN XVI:(3) pp. 58-60. (2017) Folyóiratcikk /Szakcikk /Tudományos [3215688] [ Admin láttamozott ]
2016 2. Fogarassyné Vathy Ágnes , Machalik Károly , Vassányi István , Kósa István A betegek anonimitásának biztosítása a földrajzi elhelyezkedésre kiterjedő egészségügyi adatelemzések során IME: INTERDISZCIPLINÁRIS MAGYAR EGÉSZSÉGÜGY / INFORMATIKA ÉS MENEDZSMENT AZ EGÉSZSÉGÜGYBEN XV:(1) pp. 46-50. (2016) Link(ek): Teljes dokumentum Folyóiratcikk /Szakcikk /Tudományos [3023273] [ Admin láttamozott ] 3. Rebaz A H Karim , István Vassányi , István Kósa Characterizing blood glucose response to specific meals in pre-diabetes: a small scale study In: Bari Ferenc , Almási László (szerk.) Orvosi Informatika 2016. A XXIX. Neumann Kollokvium konferencia-kiadványa . 146 p. Konferencia helye, ideje: Szeged , Magyarország , 2016.12.01 -2016.12.02. Szeged: Neumann János Számítógép-tudományi Társaság (NJSZT), 2016. pp. 93-96. (ISBN:978-963-306-514-3) Link(ek): REAL Könyvrészlet /Konferenciaközlemény /Tudományos [3149642] [ Admin láttamozott ] 4. Salai Mario , István Vassányi , István Kósa Stress Detection Using Low Cost Heart Rate Sensors JOURNAL OF HEALTHCARE ENGINEERING 2016: Paper 5136705. (2016) Link(ek): DOI, WoS Folyóiratcikk /Szakcikk /Tudományos [3092391] [ Admin láttamozott ] 5. Szálka Brigitta , Kósa István , Vassányi István , Mák Erzsébet Diabetesesek dietoterápiájának és önmenedzselésének támogatása mobilapplikációk használatával ORVOSI HETILAP 157:(29) pp. 1147-1153. (2016) Link(ek): SE Repozitórium, DOI, PubMed, MOB, WoS, Scopus Folyóiratcikk /Összefoglaló cikk /Tudományos [3092397] [ Admin láttamozott ] 6. Szálka Brigitta , Kósa István , Vassányi István , Mák Erzsébet Táplálkozási trendek, szakmai ajánlások. Mi a jövő útja?: Életmód-támogató mobil alkalmazás szerepe a dietoterápiában IME: INTERDISZCIPLINÁRIS MAGYAR EGÉSZSÉGÜGY / INFORMATIKA ÉS MENEDZSMENT AZ EGÉSZSÉGÜGYBEN 15:(7) pp. 54-58. (2016) Link(ek): MOB Folyóiratcikk /Szakcikk /Tudományos [3120434] [ Admin láttamozott ] 7. Szálka Brigitta , Vassányi István , Kállai Szilárd , Béki János , Mák Erzsébet , Kósa István mHealth szolgáltatás felhasználói igényének felmérése In: Bari Ferenc , Almási László (szerk.) Orvosi Informatika 2016. A XXIX. Neumann Kollokvium konferencia-kiadványa . 146 p. Konferencia helye, ideje: Szeged , Magyarország , 2016.12.01 -2016.12.02. Szeged: Neumann János Számítógép-tudományi Társaság (NJSZT), 2016. pp. 57-60. (ISBN:978-963-306-514-3) Link(ek): REAL Könyvrészlet /Konferenciaközlemény /Tudományos [3149634] [ Admin láttamozott ] 8. Zsolt Vassy , István Vassányi , István Kósa Changes in the spatial distribution of dominant IHD care providers over a 10 year period in Hungary In: Bari Ferenc , Almási László (szerk.) Orvosi Informatika 2016. A XXIX. Neumann Kollokvium konferencia-kiadványa . 146 p.
20
Konferencia helye, ideje: Szeged , Magyarország , 2016.12.01 -2016.12.02. Szeged: Neumann János Számítógép-tudományi Társaság (NJSZT), 2016. pp. 17-20. (ISBN:978-963-306-514-3) Link(ek): REAL Könyvrészlet /Konferenciaközlemény /Tudományos [3149638] [ Admin láttamozott ] 9. Zsolt Vassy , István Kósa , István Vassányi Stable Angina Clinical Pathway Correlation Clustering In: Rudolf Ferenc , Balázs Bánhelyi , Tamás Gergely , Attila Kertész , Zoltán Kincses (szerk.) The 10th Jubilee Conference of PhD Students in Computer Science (CS2): Volume of extended abstracts . 81 p. Konferencia helye, ideje: Szeged , Magyarország , 2016.06.27 -2016.06.29. Szeged: University of Szeged, Institute of Informatics, pp. 63-64. Befoglaló mű link(ek): Teljes dokumentum Egyéb konferenciaközlemény /Absztrakt / Kivonat /Tudományos [3168384] [ Admin láttamozott ]
2015 10. P GYUK , T LORINC , Rebaz A H K , I Vassanyi Diabetes Lifestyle Support with Improved Glycemia Prediction Algorithm In: Marike Hettinga , Åsa Smedberg , Lisette Van Gemert-Pijnen , Kari Dyb , Anne Granstrøm (szerk.) The Seventh International Conference on eHealth, Telemedicine, and Social Medicine: eTELEMED 2015 . Konferencia helye, ideje: Lisszabon , Portugália , 2015.02.22 -2015.02.27. Lisszabon: IARIA, 2015. pp. 95-100. (ISBN:978-1-61208-384-1) Könyvrészlet /Konferenciaközlemény /Tudományos [2901530] [ Admin láttamozott ] 11. Gyuk Péter , Lorincz Tamás , Rebaz A H Karim , Renner Ildikó , Vassányi István , Kósa István Vércukor-előrejelző modell klinikai validációja In: Kósa István , Vassányi István (szerk.) Új alapokon az egészségügyi informatika: A XXVIII. Neumann Kollokvium konferencia-kiadványa . 186 p. Konferencia helye, ideje: Veszprém , Magyarország , 2015.11.20 -2015.11.21. Veszprém: Neumann János Számítógéptudományi Társaság (NJSZT), 2015. pp. 96-101. (ISBN:978-615-5036-10-1) Befoglaló mű link(ek): Teljes dokumentum Könyvrészlet /Konferenciaközlemény /Tudományos [2973000] [ Admin láttamozott ] 12. Istvan Vassanyi , Istvan Kosa , Rebaz A H Karim , Marta Nemes , Brigitta Szalka Effectiveness of mobile personal dietary logging In: Piet Kommers , Pedro Isaías (szerk.) 13th International Conference on e-Society 2015 . Konferencia helye, ideje: Madeira , Portugália , 2015.03.14 -2015.03.16. Madeira: IADIS Press, 2015. pp. 288-293. (ISBN:978-989-8533-32-6) Könyvrészlet /Konferenciaközlemény /Tudományos [2901557] [ Admin láttamozott ] 13. Kósa István , Vassányi István (szerk.) Új alapokon az egészségügyi informatika: A XXVIII. Neumann Kollokvium konferencia-kiadványa Konferencia helye, ideje: Veszprém , Magyarország , 2015.11.20 -2015.11.21. Veszprém: Neumann János Számítógép-tudományi Társaság (NJSZT), 2015. 186 p. (ISBN:978-615-5036-10-1) Link(ek): Teljes dokumentum Könyv /Konferenciakötet /Tudományos [2972590] [ Szerzői rekord ] 14. Kósa István , Vassányi István , Szálka Brigitta , Nemes Márta , Cseténé Szucs Mária Fiziológiai paraméterek változása életmód támogató informatikai rendszer használata során In: Kósa István , Vassányi István (szerk.) Új alapokon az egészségügyi informatika: A XXVIII. Neumann Kollokvium konferencia-kiadványa . 186 p. Konferencia helye, ideje: Veszprém , Magyarország , 2015.11.20 -2015.11.21. Veszprém: Neumann János Számítógéptudományi Társaság (NJSZT), 2015. pp. 78-82. (ISBN:978-615-5036-10-1) Befoglaló mű link(ek): Teljes dokumentum Könyvrészlet /Konferenciaközlemény /Tudományos [2972995] [ Admin láttamozott ]
21
15. Lőrincz Tamás , Szakonyi Benedek , Gyuk Péter , Gaál Balázs , Vassányi István Flexibilis, eseményvezérelt keretrendszer mobil telemedicinális alkalmazásokhoz In: Kósa István , Vassányi István (szerk.) Új alapokon az egészségügyi informatika: A XXVIII. Neumann Kollokvium konferencia-kiadványa . 186 p. Konferencia helye, ideje: Veszprém , Magyarország , 2015.11.20 -2015.11.21. Veszprém: Neumann János Számítógéptudományi Társaság (NJSZT), 2015. pp. 153-158. (ISBN:978-615-5036-10-1) Befoglaló mű link(ek): Teljes dokumentum Könyvrészlet /Konferenciaközlemény /Tudományos [2973011] [ Admin láttamozott ] 16. Mario Salai , Istvan Vassanyi Automatic stress detection using simple telemedical heart rate meters In: Malina Jordanova , Frank Lievens (szerk.) Global Telemedicine and eHealth Updates: Knowledge Resources: Proceedings of Med-e-tel 2015 . Konferencia helye, ideje: Luxemburg , Luxemburg , 2015.04.22 -2015.04.25. Luxembourg: International Society for Telemedicine and eHealth (ISfTeH), 2015. pp. 93-97. Könyvrészlet /Konferenciaközlemény /Tudományos [2939906] [ Admin láttamozott ] 17. Mario Salai , Istvan Vassanyi Validation of a low cost telemedical stress monitoring system In: Kósa István , Vassányi István (szerk.) Új alapokon az egészségügyi informatika: A XXVIII. Neumann Kollokvium konferencia-kiadványa . 186 p. Konferencia helye, ideje: Veszprém , Magyarország , 2015.11.20 -2015.11.21. Veszprém: Neumann János Számítógéptudományi Társaság (NJSZT), 2015. pp. 180-184. (ISBN:978-615-5036-10-1) Befoglaló mű link(ek): Teljes dokumentum Könyvrészlet /Konferenciaközlemény /Tudományos [2973013] [ Admin láttamozott ] 18. Rebaz A H Karim , Vassányi István , Kósa István Hosszú hatású inzulin kezelése vércukorszint-előrejelző modellben In: Kósa István , Vassányi István (szerk.) Új alapokon az egészségügyi informatika: A XXVIII. Neumann Kollokvium konferencia-kiadványa . 186 p. Konferencia helye, ideje: Veszprém , Magyarország , 2015.11.20 -2015.11.21. Veszprém: Neumann János Számítógéptudományi Társaság (NJSZT), 2015. pp. 102-106. (ISBN:978-615-5036-10-1) Befoglaló mű link(ek): Teljes dokumentum Könyvrészlet /Konferenciaközlemény /Tudományos [2973008] [ Admin láttamozott ] 19. Rebaz A. H. Karim , Peter Gyuk , Tamas Lorincz , Istvan Vassanyi , Istvan Kosa Blood Glucose Level Prediction for Mobile Lifestyle Counseling In: Malina Jordanova , Frank Lievens (szerk.) Global Telemedicine and eHealth Updates: Knowledge Resources: Proceedings of Med-e-tel 2015 . Konferencia helye, ideje: Luxemburg , Luxemburg , 2015.04.22 -2015.04.25. Luxembourg: International Society for Telemedicine and eHealth (ISfTeH), 2015. pp. 8-12. Könyvrészlet /Konferenciaközlemény /Tudományos [2901513] [ Admin láttamozott ] 20. Szálka Brigitta , Molnár-Nemes Márta , Lorincz Tamás , Kósa István , Vassányi István , Mák Erzsébet Glikémiás hatást befolyásoló életmódbeli, étrendi tényezok vizsgálata cukorbetegek vércukor szintjére In: Kósa István , Vassányi István (szerk.) Új alapokon az egészségügyi informatika: A XXVIII. Neumann Kollokvium konferencia-kiadványa . 186 p. Konferencia helye, ideje: Veszprém , Magyarország , 2015.11.20 -2015.11.21. Veszprém: Neumann János Számítógéptudományi Társaság (NJSZT), 2015. pp. 107-111. (ISBN:978-615-5036-10-1) Befoglaló mű link(ek): Teljes dokumentum Könyvrészlet /Konferenciaközlemény /Tudományos [2973010] [ Admin láttamozott ] 21. Vassányi István , Szálka Brigitta , Nemes Márta , Gaál Balázs , Pintér Balázs Étrendértékelés és -tervezés mesterséges intelligencia segítségével In: Kósa István , Vassányi István (szerk.) Új alapokon az egészségügyi informatika: A XXVIII. Neumann Kollokvium konferencia-kiadványa . 186 p.
22
Konferencia helye, ideje: Veszprém , Magyarország , 2015.11.20 -2015.11.21. Veszprém: Neumann János Számítógéptudományi Társaság (NJSZT), 2015. pp. 83-87. (ISBN:978-615-5036-10-1) Befoglaló mű link(ek): Teljes dokumentum Könyvrészlet /Konferenciaközlemény /Tudományos [2972998] [ Admin láttamozott ] 22. Vassy Zsolt , Kósa István , Vassányi István Stabil anginás betegutak klaszterelemzése In: Kósa István , Vassányi István (szerk.) Új alapokon az egészségügyi informatika: A XXVIII. Neumann Kollokvium konferencia-kiadványa . 186 p. Konferencia helye, ideje: Veszprém , Magyarország , 2015.11.20 -2015.11.21. Veszprém: Neumann János Számítógéptudományi Társaság (NJSZT), 2015. pp. 7-10. (ISBN:978-615-5036-10-1) Befoglaló mű link(ek): Teljes dokumentum Könyvrészlet /Konferenciaközlemény /Tudományos [2972991] [ Admin láttamozott ] Független idéző: 1 Összesen: 1 1.
Machalik Károly, Fogarassyné Vathy Ágnes Betegségek kapcsolatrendszerének hálózatai In: XXIX. Neumann Kollokvium. Szeged, Magyarország: 2016.12.01-2016.12.02. NJSzT, (2016.) , pp. 11-15 . ISBN: 978-963-306-514-3 (NJSzT)
Egyéb konferenciaközlemény /Konferenciaközlemény /Tudományos [16249892]
2014 23. Attila Nemes , Ferenc Király , István Vassányi , István Kósa The impact of geographical distances to coronary angiography laboratories on the patient evaluation pathways in patients with suspected coronary artery disease. Results from a population-based study in Hungary POSTEPY W KARDIOLOGII INTERWENCYJNEJ 10:(4) pp. 270-273. (2014) Link(ek): DOI, PubMed, WoS, Scopus, Teljes dokumentum Folyóiratcikk /Szakcikk /Tudományos [2836910] [ Szerzői rekord ] Független idéző: 1 Összesen: 1 1
M B Faslur Rahuman, Jayanthimala B Jayawardena, George R Francis A comparison of rescue and primary percutaneous coronary interventions for acute ST elevation myocardial infarction INDIAN HEART JOURNAL (ISSN: 0019-4832) 69: Paper 10.1016/j.ihj.2017.02.019. (2017) Link(ek): DOI
Folyóiratcikk [16514262]
24. Béres Mónika , Vassányi István , Fabio Babiloni , Giovanni Vecchiato Képfeldolgozás folyamata az EEG neuroesztétikai vizsgálatában IME: INTERDISZCIPLINÁRIS MAGYAR EGÉSZSÉGÜGY / INFORMATIKA ÉS MENEDZSMENT AZ EGÉSZSÉGÜGYBEN XIII:(Képalkotó különszám) pp. 25-28. (2014) Folyóiratcikk /Szakcikk /Tudományos [2836530] [ Szerzői rekord ] 25. Ferenc Király , István Kósa , István Vassányi The Effect of the Waiting Times on the Patient Pathways for Patients with Suspected Coronary Artery Disease In: Lăcrămioara Stoicu-Tivadar , Simon de Lusignan , Andrej Orel , Rolf Engelbrecht , György Surján (szerk.) Cross-Border Challenges in Informatics with a Focus on Disease Surveillance and Utilising Big Data: EFMI Special Topic Conference . Konferencia helye, ideje: Budapest; Barcelona , Magyarország , 2014.04.26 -2014.04.29. Amsterdam: IOS Press, 2014. pp. 97-101. (ISBN:978-1-61499-388-9) Könyvrészlet /Konferenciaközlemény /Tudományos [2836749] [ Szerzői rekord ] 26. Gyuk Péter , Vassányi István , Kósa István Diabétesz életmód-támogatás vércukorszint-előrejelzéssel IME: INTERDISZCIPLINÁRIS MAGYAR EGÉSZSÉGÜGY / INFORMATIKA ÉS MENEDZSMENT AZ EGÉSZSÉGÜGYBEN 13:(4) pp. 44-47. (2014) Link(ek): Teljes dokumentum, Matarka Folyóiratcikk /Szakcikk /Tudományos [2769425] [ Admin láttamozott ] 27. István Kósa , István Vassányi , Balázs Pintér , Márta Nemes , Krisztina Kámánné , László Kohut Preclinical tests of an android based dietary logging application In: Lăcrămioara Stoicu-Tivadar , Simon de Lusignan , Andrej Orel , Rolf Engelbrecht , György Surján (szerk.) Cross-Border Challenges in Informatics with a Focus on Disease Surveillance and Utilising Big Data: EFMI Special Topic Conference . Konferencia helye, ideje: Budapest; Barcelona , Magyarország , 2014.04.26 -2014.04.29. Amsterdam: IOS
23
Press, 2014. pp. 53-57. (ISBN:978-1-61499-388-9) Könyvrészlet /Konferenciaközlemény /Tudományos [2836743] [ Szerzői rekord ] 28. István VASSÁNYI , István KÓSA , Balázs PINTÉR , Balázs GAÁL Personalized Dietary Counseling System Using Harmony Rules in Tele-Care JOURNAL OF BIOMEDICAL INFORMATICS 10:(2) pp. 17-22. (2014) Link(ek): Teljes dokumentum Folyóiratcikk /Szakcikk /Tudományos [2836710] [ Admin láttamozott ] 29. Király Ferenc , Nemes Attila , Vassányi István , Kósa István A szívkatéteres laboratóriumoktól mért földrajzi távolság hatása az iszkémiás szívbetegség gyanújával ellátásra került betegek ellátási útjára IME: INTERDISZCIPLINÁRIS MAGYAR EGÉSZSÉGÜGY / INFORMATIKA ÉS MENEDZSMENT AZ EGÉSZSÉGÜGYBEN 13:(3) pp. 15-18. (2014) Link(ek): Teljes dokumentum, Matarka Folyóiratcikk /Szakcikk /Tudományos [2769426] [ Admin láttamozott ] Független idéző: 1 Összesen: 1 1.
Machalik Károly, Fogarassyné Vathy Ágnes Betegségek kapcsolatrendszerének hálózatai In: XXIX. Neumann Kollokvium. Szeged, Magyarország: 2016.12.01-2016.12.02. Szeged: NJSzT, (2016.) , pp. 11-15 . ISBN: 978-963-306-514-3 (NJSzT)
Egyéb konferenciaközlemény /Konferenciaközlemény /Tudományos [16249899]
30. I Kósa , I Vassányi , M Nemes , K H Kálmánné , B Pintér , L Kohut A fast, android based dietary logging application to support the life style change of cardio-metabolic patients In: Malina Jordanova , Frank Lievens (szerk.) Global Telemedicine and eHealth Updates: Knowledge Resources: Med-e-Tel Conference 2014 . Konferencia helye, ideje: Luxembourg , Luxemburg , 2014.04.09 -2014.04.11. pp. 553-556. Egyéb konferenciaközlemény /Konferenciaközlemény /Tudományos [2836702] [ Szerzői rekord ] 31. Kósa István , Király Ferenc , Vassányi István , Simon Attila , Gödölle Zoltán , Merth Gabriella , Kohut László Rehabilitációra érdemes és ténylegesen rehabilitációra kerülo betegellátási utak elemzése In: Bari Ferenc , Almási László (szerk.) Orvosi Informatika 2014: A XXVII. Neumann Kollokvium konferencia-kiadványa . 162 p. Konferencia helye, ideje: Szeged , Magyarország , 2014.11.21 -2014.11.22. Veszprém: Pannon Egyetem, 2014. pp. 9-13. (ISBN:978-963-396-040-0) Befoglaló mű link(ek): Egyéb URL Könyvrészlet /Konferenciaközlemény /Tudományos [2836521] [ Szerzői rekord ] 32. Kozmann G , Tuboly G , Vassányi I , Szathmáry V Kamrai szívizom-repolarizáció heterogenitás vizsgálat bioelektromos képalkotóval IME: INTERDISZCIPLINÁRIS MAGYAR EGÉSZSÉGÜGY / INFORMATIKA ÉS MENEDZSMENT AZ EGÉSZSÉGÜGYBEN XIII.évf.:(3.szám) pp. 47-51. (2014) Link(ek): Teljes dokumentum Folyóiratcikk /Szakcikk /Tudományos [2722148] [ Admin láttamozott ] 33. Kozmann György , Tuboly Gergely , Vassányi István Kamrai szívizom-repolarizáció heterogenitás modellezéses vizsgálata IME: INTERDISZCIPLINÁRIS MAGYAR EGÉSZSÉGÜGY / INFORMATIKA ÉS MENEDZSMENT AZ EGÉSZSÉGÜGYBEN 13:(Képalkotó különszám) pp. 31-34. (2014) Link(ek): Teljes dokumentum Folyóiratcikk /Szakcikk /Tudományos [2805135] [ Admin láttamozott ] 34. Mario Salai , Gergely Tuboly , István Vassányi , István Kósa Telemedical Heart Rate Measurements for Lifestyle Counselling HUNGARIAN JOURNAL OF INDUSTRY AND CHEMISTRY 42:(2) pp. 73-78. (2014) 13th International PhD Workshop on Systems and Control Conference. Veszprém, Magyarország: 2014.08.25 Link(ek): Teljes dokumentum Folyóiratcikk /Szakcikk /Tudományos [2773361] [ Admin láttamozott ]
24
35. Nemes Márta , Vassányi istván , Kósa István , Pintér Balázs Okostelefon alapú diéta-naplózó rendszer diabeteses betegek számára ÚJ DIÉTA: A MAGYAR DIETETIKUSOK LAPJA (2001-) 2014:(5) pp. 15-17. (2014) Folyóiratcikk /Szakcikk /Tudományos [2836475] [ Admin láttamozott ] 36. Nemes Márta , Vassányi István , Kámánné Hózler Krisztina , Pintér Balázs , Kohut László , Kósa István Életmód-naplózás pontosságának elemzése kardiológiai rehabilitációra kerülo betegek körében In: Bari Ferenc , Almási László (szerk.) Orvosi Informatika 2014: A XXVII. Neumann Kollokvium konferencia-kiadványa . 162 p. Konferencia helye, ideje: Szeged , Magyarország , 2014.11.21 -2014.11.22. Veszprém: Pannon Egyetem, 2014. pp. 5-9. (ISBN:978-963-396-040-0) Befoglaló mű link(ek): Egyéb URL Könyvrészlet /Konferenciaközlemény /Tudományos [2836506] [ Szerzői rekord ] 37. Peter Gyuk , Istvan Szabo , Istvan Vassanyi , Istvan Kosa , Levente Kovacs Combined Model for Diabetes Lifestyle Support In: Lăcrămioara Stoicu-Tivadar , Simon de Lusignan , Andrej Orel , Rolf Engelbrecht , György Surján (szerk.) Cross-Border Challenges in Informatics with a Focus on Disease Surveillance and Utilising Big Data: EFMI Special Topic Conference . Konferencia helye, ideje: Budapest; Barcelona , Magyarország , 2014.04.26 -2014.04.29. Amsterdam: IOS Press, 2014. pp. 77-81. (ISBN:978-1-61499-388-9) Könyvrészlet /Konferenciaközlemény /Tudományos [2836746] [ Szerzői rekord ] 38. I Szabó , P Gyuk , I Vassányi Lifestyle Log Based Blood Glucose Level Prediction for Outpatient Care In: eTELEMED 2014, The Sixth International Conference on eHealth, Telemedicine, and Social Medicine . Konferencia helye, ideje: Barcelona , Spanyolország , 2014.03.23 -2014.03.25. 2014. pp. 205-210. (ISBN:978-1-61208-327-8) Könyvrészlet /Konferenciaközlemény /Tudományos [2836731] [ Szerzői rekord ] 39. Szalai Márió , Tuboly Gergely , Vassányi István , Kósa István Reliability of telemedical Heart Rate meters IME: INTERDISZCIPLINÁRIS MAGYAR EGÉSZSÉGÜGY / INFORMATIKA ÉS MENEDZSMENT AZ EGÉSZSÉGÜGYBEN 13:(5) pp. 49-55. (2014) Link(ek): Teljes dokumentum Folyóiratcikk /Szakcikk /Tudományos [2733717] [ Admin láttamozott ] 40. Tamás Réka , Cseténé Szucs Mária , Nemes Márta , Vassányi István , Pintér Balázs , Kohut László , Kósa István Felhasználói tapasztalatok a Lavinia mobil életmód-tükör alkalmazással In: Bari Ferenc , Almási László (szerk.) Orvosi Informatika 2014: A XXVII. Neumann Kollokvium konferencia-kiadványa . 162 p. Konferencia helye, ideje: Szeged , Magyarország , 2014.11.21 -2014.11.22. Veszprém: Pannon Egyetem, 2014. pp. 13-17. (ISBN:978-963-396-040-0) Befoglaló mű link(ek): Egyéb URL Könyvrészlet /Konferenciaközlemény /Tudományos [2836515] [ Szerzői rekord ] 41. Tuboly G , Kozmann G , Vassányi I Távmonitorozásra is alkalmas pitvari fibrilláció detektálási módszer IME: INTERDISZCIPLINÁRIS MAGYAR EGÉSZSÉGÜGY / INFORMATIKA ÉS MENEDZSMENT AZ EGÉSZSÉGÜGYBEN XIII.évf:(1.szám) pp. 51-54. (2014) Link(ek): Teljes dokumentum Folyóiratcikk /Szakcikk /Tudományos [2722146] [ Admin láttamozott ]
2013 42. Balázs PINTÉR , Balázs GAÁL , István VASSÁNYI Formalizing harmony rules for nutrition counseling EGYPTIAN COMPUTER SCIENCE JOURNAL 37:(7) pp. 24-28. (2013) Int. Conf. on Euro-Mediterranean Medical Informatics and Telemedicine, EMMIT'2013.. Marokkó: 2013.10.21 -2013.10.23.
25
Folyóiratcikk /Szakcikk /Tudományos [2836725] [ Szerzői rekord ] 43. P Cserti , Sz Szondi , B Gaal , I Vassanyi GAME: GPU accelerated multipurpose evolutionary algorithm library International Journal of Innovative Computing and Applications 5:(3) (2013) Link(ek): DOI Folyóiratcikk /Szakcikk /Tudományos [2442468] [ Szerzői rekord ] Független idéző: 2 Összesen: 2 1
Guo-yan Meng, Yu-lan Hu, Yun Tian, Qing-Shan Zhao Adaptive bacterial colony chemotaxis multi-objective optimisation algorithm INTERNATIONAL JOURNAL OF COMPUTING SCIENCE AND MATHEMATICS (ISSN: 1752-5055) 5: (4) pp. 336-345. (2014) Link(ek): DOI
Folyóiratcikk [15983917] 2
Youssef Safi, Abdelaziz Bouroumi Evolutionary single hidden-layer feed forward networks International Journal of Innovative Computing and Applications 6: (2) pp. 73-86. (2015) Link(ek): DOI
Folyóiratcikk /Szakcikk /Tudományos [15983929]
44. Z Juhasz , I Vassanyi , A G Nagy , A Papp , D Fabo , Gy Kozmann SOLO: An EEG Processing Software Framework for Localising Epileptogenic Zones In: Ján Manka , Milan Tysler , Viktor Witkovsky , Ivan Frollo (szerk.) Measurement 2013: 9th International Conference on Measurement . Konferencia helye, ideje: Smolenice , Szlovákia , 2013.05.27 -2013.05.30. Bratislava: Vydavatelstvo Slovenskej Akadémie Vied (VEDA), 2013. pp. 105-108. (ISBN:978-80-969-672-5-4) Könyvrészlet /Konferenciaközlemény /Tudományos [2442447] [ Admin láttamozott ] 45. Karikó Ivett , Vassányi István Táplálkozási adatbázisok átjárhatósága In: Kósa István , Vassányi István (szerk.) Az e-Health kihívásai. A XXVI. Neumann Kollokvium kiadványa . 210 p. Konferencia helye, ideje: Veszprém , Magyarország , 2013.11.22 -2013.11.23. Veszprém: Pannon Egyetem, 2013. pp. 103-106. (ISBN:978-615-5044-90-8) Befoglaló mű link(ek): OSZK Könyvrészlet /Konferenciaközlemény /Tudományos [2507466] [ Admin láttamozott ] 46. Király F , Vassányi I , Nemes A , Kósa I REGIONÁLIS ELTÉRÉSEK A KORONAROGRÁFIA, MINT KIINDULÓ VIZSGÁLAT ALKALMAZÁSA TEKINTETÉBEN KOSZORÚÉR-BETEGSÉG GYANÚ MIATT KIVIZSGÁLÁSRA KERÜLT BETEGEKBEN CARDIOLOGIA HUNGARICA 43: p. B71. (2013) Folyóiratcikk /Absztrakt / Kivonat /Tudományos [2527153] [ Szerzői rekord ] 47. Király Ferenc , Vassányi István , Rárosi Ferenc , Nemes Attila , Kósa István A regionális vizsgálati frekvenciák és a viszgálatra kerülok halálozási mutatójának összefüggése koszorúérbetegség gyanújával kivizsgálásra kerülo betegekben In: Kósa István , Vassányi István (szerk.) Az e-Health kihívásai. A XXVI. Neumann Kollokvium kiadványa . 210 p. Konferencia helye, ideje: Veszprém , Magyarország , 2013.11.22 -2013.11.23. Veszprém: Pannon Egyetem, 2013. pp. 135-138. (ISBN:978-615-5044-90-8) Befoglaló mű link(ek): OSZK Könyvrészlet /Konferenciaközlemény /Tudományos [2507463] [ Admin láttamozott ] 48. Kósa I , Nemes A , Belicza E , Király F , Vassányi I Regional differences in the utilisation of coronary angiography as initial investigation for the evaluation of patients with suspected coronary artery disease INTERNATIONAL JOURNAL OF CARDIOLOGY 168:(5) pp. 5012-5015. (2013) Link(ek): DOI, PubMed, WoS, Scopus Folyóiratcikk /Szakcikk /Tudományos [2376933] [ Hitelesített ] Független idéző: 1 Összesen: 1 1
Imbalzano E, Ceravolo R, Di Stefano R, Vatrano M, Saitta A
26
Rare combination of left ventricular noncompaction, bicuspid aortic valve and myocardial bridging. Rare case or common genetic mutations? INTERNATIONAL JOURNAL OF CARDIOLOGY (ISSN: 0167-5273) 171: (3) pp. E90-E92. (2014) Link(ek): DOI, WoS, Scopus
Folyóiratcikk /Rövid közlemény /Tudományos [13738611]
49. Kósa I , Király F , Nemes A , Vassányi I AZ INVAZÍV SZOLGÁLTATÓ TÁVOLSÁGÁNAK HATÁSA A BETEGUTAKRA KOSZORÚÉRBETEGSÉG GYANÚJA MIATTI KIVIZSGÁLÁSOK SORÁN CARDIOLOGIA HUNGARICA 43: p. B44. (2013) Folyóiratcikk /Absztrakt / Kivonat /Tudományos [2527136] [ Admin láttamozott ] 50. Kósa István , Tamás Réka , Vassányi István , Nemes Márta , Kozmann György Életmód-változtatást támogató mobil informatikai alkalmazások IME: INTERDISZCIPLINÁRIS MAGYAR EGÉSZSÉGÜGY / INFORMATIKA ÉS MENEDZSMENT AZ EGÉSZSÉGÜGYBEN 12:(8) pp. 58-62. (2013) Folyóiratcikk /Szakcikk /Tudományos [2442429] [ Admin láttamozott ] 51. Kósa István , Vassányi István (szerk.) Az e-Health kihívásai. A XXVI. Neumann Kollokvium kiadványa Konferencia helye, ideje: Veszprém , Magyarország , 2013.11.22 -2013.11.23. Veszprém: Pannon Egyetem, 2013. 210 p. (ISBN:978-615-5044-90-8) Link(ek): OSZK Könyv /Konferenciakötet /Tudományos [2507408] [ Szerzői rekord ] 52. Kósa István , Merth Gabriella , Rárosi Ferenc , Vassányi István , Kohut László Kardiológiai rehabilitációs kezelésben részesülo illetve ilyen kezelésre potenciálisan jelölt betegek gyógyszerfogyasztásának összehasonlítása In: Kósa István , Vassányi István (szerk.) Az e-Health kihívásai. A XXVI. Neumann Kollokvium kiadványa . 210 p. Konferencia helye, ideje: Veszprém , Magyarország , 2013.11.22 -2013.11.23. Veszprém: Pannon Egyetem, 2013. pp. 147-150. (ISBN:978-615-5044-90-8) Befoglaló mű link(ek): OSZK Könyvrészlet /Konferenciaközlemény /Tudományos [2507440] [ Admin láttamozott ] 53. Kozmann Gy , Juhász Z , Tuboly G , Vassányi I , Nagy Z Lokális agykérgi aktivitás mérése Laplace-típusú EEG térképezéssel: A felbontás vizsgálata modellezéssel IME: INTERDISZCIPLINÁRIS MAGYAR EGÉSZSÉGÜGY / INFORMATIKA ÉS MENEDZSMENT AZ EGÉSZSÉGÜGYBEN 12:(Képalkotó különszám) pp. 24-28. (2013) Link(ek): Teljes dokumentum Folyóiratcikk /Szakcikk /Tudományos [2508430] [ Szerzői rekord ] 54. A Lukosevicius (szerk.) Biomedical Engineering 2013: Proceedings of International Conference Konferencia helye, ideje: Kaunas , Litvánia , 2013.11.28 -2013.11.29. Kaunas: Kaunasz Akademija, 2013. Könyv /Konferenciakötet /Tudományos [2507533] [ Admin láttamozott ] 55. Nemes Márta , Vassányi István , Tamás Réka , Kósa István Étel adatbázisok tartalmi eltéréseinek hatása a diéta naplózás pontosságára In: Kósa István , Vassányi István (szerk.) Az e-Health kihívásai. A XXVI. Neumann Kollokvium kiadványa . 210 p. Konferencia helye, ideje: Veszprém , Magyarország , 2013.11.22 -2013.11.23. Veszprém: Pannon Egyetem, 2013. pp. 91-94. (ISBN:978-615-5044-90-8) Befoglaló mű link(ek): OSZK Könyvrészlet /Konferenciaközlemény /Tudományos [2507477] [ Admin láttamozott ]
27
56. Pintér Balázs , Gaál Balázs , Vassányi István Személyre szabott táplálkozás-tanácsadó rendszer harmónia szabályokkal In: Kósa István , Vassányi István (szerk.) Az e-Health kihívásai. A XXVI. Neumann Kollokvium kiadványa . 210 p. Konferencia helye, ideje: Veszprém , Magyarország , 2013.11.22 -2013.11.23. Veszprém: Pannon Egyetem, 2013. pp. 99-102. (ISBN:978-615-5044-90-8) Befoglaló mű link(ek): OSZK Könyvrészlet /Konferenciaközlemény /Tudományos [2507472] [ Admin láttamozott ] 57. I Vassányi , B Pintér , M Nemes , I Kósa , E Mák , Gy Kozmann Dietary Logging and Analysis for Tele-Care Using Harmony Rules In: Malina Jordanova , Frank Lievens (szerk.) Global Telemedicine and eHealth Updates: Knowledge Resources: Proceedings of Med-e-tel 2013 . Konferencia helye, ideje: Luxembourg , Luxemburg , 2013.04.10 -2013.04.12. Luxembourg: Future Medicine Ltd., 2013. pp. 379-383. (ISBN:ISSN 1998-5509) Könyvrészlet /Konferenciaközlemény /Tudományos [2442457] [ Admin láttamozott ] 58. I Vassanyi , I Kosa , B Pinter Challenges and limits in personalized dietary logging and analysis In: A Lukosevicius (szerk.) Biomedical Engineering 2013: Proceedings of International Conference . Konferencia helye, ideje: Kaunas , Litvánia , 2013.11.28 -2013.11.29. Kaunas: Kaunasz Akademija, 2013. pp. 22-25. Könyvrészlet /Konferenciaközlemény /Tudományos [2507534] [ Admin láttamozott ] 59. Vassányi István , Juhász Zoltán , Kozmann György , Fabó Dániel Epileptikus gócok EEG alapú lokalizálását támogató szoftver környezet IME: INTERDISZCIPLINÁRIS MAGYAR EGÉSZSÉGÜGY / INFORMATIKA ÉS MENEDZSMENT AZ EGÉSZSÉGÜGYBEN XII:(4) pp. 43-46. (2013) Folyóiratcikk /Szakcikk /Tudományos [2378504] [ Admin láttamozott ]
2012 60. B Pinter , I Vassányi , B Gaál , E Mák , Gy Kozmann Personalized Nutrition Counseling Expert System In: Jobbágy Á (szerk.) Proceedings of the 5th European Conference of the International Federation for Medical and Biological Engineering . Konferencia helye, ideje: Budapest , Magyarország , 2011.09.14 -2011.09.18. (IFMBE International Federation for Medical and Biological Engineering) Berlin; Heidelberg: Springer Verlag, 2012. pp. 957-960. ( IFMBE Proceedings; 37. ) (ISBN:978-3-642-23508-5) Link(ek): DOI Befoglaló mű link(ek): DOI Könyvrészlet /Konferenciaközlemény /Tudományos [1881820] [ Admin láttamozott ] Független idéző: 1 Összesen: 1 1
Dalia I Hemdan An Automated Balanced Nutritional Guidance system Based on Rough Sets International Journal of Computer and Information Technology (IJCIT) (ISSN: 2279-0764) 5: (2) pp. 187-193. (2016) Folyóiratcikk /Szakcikk /Tudományos [15983957]
61. Cserti P , Sz.Szondi , B.Gaál , Gy.Kozmann , I.Vassányi GPU based parallel genetic algorithm library In: Filipic B , Silc J (szerk.) Int. Conf. on Bioinspired Optimization Methods and their Applications - BIOMA 2012 . Konferencia helye, ideje: Ljubljana , Szlovénia , 2012.05.24 -2012.05.25. Ljubljana: Jozef Stefan Institute, 2012. pp. 231-245. (ISBN:978-961-264-043-9) Könyvrészlet /Konferenciaközlemény /Tudományos [2180782] [ Admin láttamozott ] 62. B Pintér , I.Vassányi , B.Gaál , Gy.Kozmann MenuGene: A Comprehensive Expert System for Dietary and Lifestyle Counseling and Tracking In: Y Papadopoulos (szerk.)
28
8th International Conference on Information Technology and Computer Science . Konferencia helye, ideje: Athens , Görögország , 2012.05.21 -2012.05.24. Athens: Athens Institute for Education and Research, 2012. pp. 257-268. (ISBN:978-960-9549-60-8) Könyvrészlet /Konferenciaközlemény /Tudományos [2180784] [ Admin láttamozott ] 63. Pinter B , Vassanyi I , Gaal B , Mak E , Kozmann G Personalized Nutrition Counseling Expert System IFMBE PROCEEDINGS 37: pp. 957-960. (2012) Link(ek): WoS, Teljes dokumentum Folyóiratcikk /Konferenciaközlemény /Tudományos [2194373] [ Admin láttamozott ] 64. Vassányi István , Kozmann György , Kósa István , Nemes Attila , Hortobágyi József Egészségügyi adatvagyon hasznosítása a stabil coronaria betegek ellátásának elemzésére IME: INTERDISZCIPLINÁRIS MAGYAR EGÉSZSÉGÜGY / INFORMATIKA ÉS MENEDZSMENT AZ EGÉSZSÉGÜGYBEN 11:(4) pp. 41-45. (2012) Link(ek): Teljes dokumentum, Egyéb URL, Matarka Folyóiratcikk /Szakcikk /Tudományos [2121528] [ Szerzői rekord ]
2011 65. Gaál Balázs , Vassányi István , Pintér Balázs , Kozmann György , Mák Erzsébet Táplálkozás-tanácsadó szakértői rendszer IME: INTERDISZCIPLINÁRIS MAGYAR EGÉSZSÉGÜGY / INFORMATIKA ÉS MENEDZSMENT AZ EGÉSZSÉGÜGYBEN 10:(5) pp. 33-37. (2011) Link(ek): Teljes dokumentum Folyóiratcikk /Szakcikk /Tudományos [1881779] [ Admin láttamozott ] Független idéző: 1 Összesen: 1 1
Pintér Patricia Cukorbetegek étrendtervezését támogató otthoni felügyeleti rendszer IME: INTERDISZCIPLINÁRIS MAGYAR EGÉSZSÉGÜGY / INFORMATIKA ÉS MENEDZSMENT AZ EGÉSZSÉGÜGYBEN (ISSN: 1588-6387) XI: (4) pp. 49-51. (2011) Folyóiratcikk /Szakcikk /Tudományos [15972698]
66. István Vassányi , György Kozmann , András Bánhalmi , Balázs Végsö , István Kósa , Tibor Dulai , Zsolt Tarjányi , Gergely Tuboly , Péter Cserti , Balázs Pintér Applications of Medical Intelligence in Remote Monitoring STUDIES IN HEALTH TECHNOLOGY AND INFORMATICS 169: pp. 671-675. (2011) Link(ek): DOI, PubMed, Scopus Folyóiratcikk /Szakcikk /Tudományos [1881486] [ Admin láttamozott ] 67. István Vassányi , György Kozmann , András Bánhalmi , Balázs Végso , István Kósa , Tibor Dulai , Zsolt Tarjányi , Gergely Tuboly , Péter Cserti , Balázs Pintér Applications of medical intelligence in remote monitoring In: A Moen (szerk.) XXIII International Conference of the European Federation for Medical Informatics . Konferencia helye, ideje: Oslo , Norvégia , 2011.08.28 -2011.08.31. Oslo: IOS Press, pp. 671-675. ( Studies in Health Technology and Informatics ; 169. ) User Centered Networked Health Care Egyéb konferenciaközlemény /Konferenciaközlemény /Tudományos [3128731] [ Szerzői rekord ] 68. I Kósa , B Végso , I Vassányi , Zs Tarjányi , T Dulai , Gy Kozmann , Cs Csoma , A.Davies Tele-monitoring for neurological patients: lessons learned In: M Jordanova (szerk.) Global Telemedicine and eHealth Updates: Knowledge Resources . Konferencia helye, ideje: Luxembourg , Luxemburg , 2011.04.01 -2011.04.03. Luxexpo, pp. 602-205. 4 Egyéb konferenciaközlemény /Konferenciaközlemény /Tudományos [3128735] [ Szerzői rekord ] 69. Kósa I , Vassányi I , Nemes A , Hortobágyi J , Kozmann G Stress ECG utilization in the evaluation of patients with chest pain: The real practice in Hungary with 10 million inhabitants INTERNATIONAL JOURNAL OF CARDIOLOGY 149:(1) pp. 137-139. (2011) Link(ek): DOI, PubMed, WoS, Scopus Folyóiratcikk /Szakcikk /Tudományos [2042200]
29
[ Admin láttamozott ] 70. Kosa I , Vassanyi I , Egyed CS , Kozmann GY The role of nuclear cardiology in the evaluation of patients before coronary angiography EUROPEAN HEART JOURNAL SUPPLEMENTS 13:(A) p. A14. 1 p. (2011) Link(ek): WoS Folyóiratcikk /Absztrakt / Kivonat /Tudományos [2043674] [ Admin láttamozott ] 71. Lohner Roland , Szűcs Vilmos , Vassányi István , Bilicki Vilmos , Dévényi Csaba , Dulai Tibor , Kósa István Remote health monitoring system Lajstromszám: 11462017.2-2319 Benyújtás éve: 2011. Közzététel éve: 2011 Benyújtás helye: Németország Oltalmi formák /Európai szabadalom /Tudományos [1895003] [ Admin láttamozott ] 72. Vassányi István , Végső Balázs , Dulai Tibor , Kozmann György , Szűcs Vilmos , Kósa István Orvosi intelligencia alkalmazásai a távdiagnosztikában IME: INTERDISZCIPLINÁRIS MAGYAR EGÉSZSÉGÜGY / INFORMATIKA ÉS MENEDZSMENT AZ EGÉSZSÉGÜGYBEN 10:(9) pp. 51-53. (2011) Link(ek): Teljes dokumentum Folyóiratcikk /Szakcikk /Tudományos [2036766] [ Admin láttamozott ]
2010 73. Colombo G , Merico D , Boncoraglio G , De Paoli F , Ellul J , Frisoni G , Nagy Z , van der Lugt A , Vassanyi I , Antoniotti M An ontological modeling approach to cerebrovascular disease studies: the NEUROWEB case. JOURNAL OF BIOMEDICAL INFORMATICS 43:(4) pp. 469-484. (2010) Link(ek): DOI, PubMed, WoS Folyóiratcikk /Szakcikk /Tudományos [1625080] [ Szerzői rekord ] Független idéző: 15 Összesen: 15 1
Hyeoneui Kim, Jeeyae Choi, Lelanie Secalag, Laura Dibsie, Aziz Boxwala, Lucila Ohno-Machado Building an Ontology for Pressure Ulcer Risk Assessment to Allow Data Sharing and Comparisons Across Hospitals AMIA Annual Symposium Proceedings (ISSN: 1559-4076) 2010: pp. 382-386. (2010) Link(ek): Pubmed Central
Folyóiratcikk /Szakcikk /Tudományos [15972562] 2.
Olsson C E, Kemp G J L Standardizing radiation oncology data for future modelling of side effects after radiation therapy In: , Proceedings. Glasgow, Skócia: 2011.10.24-2011.10.28. (2011.) , pp. 67-70 . ISBN: 9781450309547 Link(ek): DOI, Scopus
Egyéb konferenciaközlemény [12026382] 3.
Nathalia Fernanda Gómez González Apoyo para el diagnóstico de enfermedades vasculares - ADEV Témavezető(k): Correo Javeriano. 37 p. 2012.
Disszertáció /PhD /Tudományos [15972694] 4
C Machado Toward a Translational Medicine Approach for Hypertrophic Cardiomyopathy In: C Böhm (szerk.) : InformationTechnology in Bio- and Medical Informatics. (7451) Berlin: Springer, 2012. (ISBN 978-3-642-32394-2) pp. 151-165. (Lecture Notes in Computer Science 7451)
Könyvrészlet [12917470] 5
Maramis C, Falelakis M, Lekka I, Diou C, Mitkas P, Delopoulos A Applying semantic technologies in cervical cancer research Data and Knowledge Engineering 86: pp. 160-178. (2013) Link(ek): DOI, WoS, Scopus
Folyóiratcikk [13330087] 6.
Cátia Maria Machado Knowledge Representation for Data Integration and Exploration in Translational Medicine Témavezető(k): Francisco Moreira Couto. 163 p. 2013.
Disszertáció /PhD /Tudományos [15972624] 7
Liaw ST, Rahimi A, Ray P, Taggart J, Dennis S, de Lusignan S, Jalaludin B, Yeo AET, Talaei-Khoei A Towards an ontology for data quality in integrated chronic disease management: A realist review of the literature INTERNATIONAL JOURNAL OF MEDICAL INFORMATICS (ISSN: 1386-5056) 82: (1) pp. 10-24. (2013) Link(ek): DOI, PubMed, WoS, Scopus
Folyóiratcikk [13330086] 8
Alireza Rahimi, Nandan Parameswaran, Pradeep Kumar Ray Development of a Methodological Approach for Data Quality Ontology in Diabetes Management International Journal of E-Health and Medical Communications (ISSN: 1947-315X) 5: (3) p. 1. 20 p. (2014) Link(ek): DOI
Folyóiratcikk /Szakcikk /Tudományos [15972666]
30
9
Teresa Podsiadly-Marczykowska, Bogdan Ciszek, Artur Przelaskowski Development of Diagnostic Stroke Ontology - Preliminary Results In: Information Technologies in Biomedicine. (Volume 4) Springer, 2014. (ISBN 978-3-319-06595-3) pp. 261-272. (Advances in Intelligent Systems and Computing) Link(ek): DOI
Könyvrészlet /Könyvfejezet /Tudományos [15972611] 10
Siaw-Teng Liaw, Jane Taggart, Hairong Yu, Alireza Rahimi Electronic health records and disease registries to support integrated care in a health neighbourhood: an ontology-based methodology AMIA Summits on Translational Science Proceedings 2014: pp. 50-54. (2014) Link(ek): Pubmed Central
Folyóiratcikk /Szakcikk /Tudományos [15972679] 11
Liaw Siaw-Teng, Taggart Jane, Yu Hairong, de Lusignan Simon, Kuziemsky Craig, Hayen Andrew Integrating electronic health record information to support integrated care: Practical application of ontologies to improve the accuracy of diabetes disease registers JOURNAL OF BIOMEDICAL INFORMATICS (ISSN: 1532-0464) 52: pp. 364-372. (2014) Link(ek): DOI, WoS
Folyóiratcikk /Szakcikk /Tudományos [14784783] 12.
Susana Rodrigues Santos, Ana Teresa Freitas, Alexandra Fernandes Overview of Hypertrophic Cardiomyopathy (HCM) Genomics and Transcriptomics: Molecular Tools in HCM Assessment for Applic ation in Clinical 28 p. Universidade de Lisboa. https://sites.fct.unl.pt/sites/default/files/human-genetics-and-cancer-therapeutics-at-fct/files /overview_of_hypertrophic_cardiomyopathy.pdf. (2014.)
Egyéb /Kutatási jelentés (közzétett) /Tudományos [15972574] 13
Rahimi A, Liaw ST, Taggart J, Ray P, Yu H Validating an ontology-based algorithm to identify patients with type 2 diabetes mellitus in electronic health records INTERNATIONAL JOURNAL OF MEDICAL INFORMATICS (ISSN: 1386-5056) 83: (10) pp. 768-778. (2014) Link(ek): DOI, WoS
Folyóiratcikk /Szakcikk /Tudományos [14508818] 14
Machado Catia M, Rebholz-Schuhmann Dietrich, Freitas Anat, Couto Francisco M The semantic web in translational medicine: current applications and future directions BRIEFINGS IN BIOINFORMATICS (ISSN: 1467-5463) 16: (1) pp. 89-103. (2015) Link(ek): DOI, WoS
Folyóiratcikk /Szakcikk /Tudományos [14784782] 15
SO Danso, DE Job, DR Gonzalez, DA Dickie, J Palmer Developing an Integrated Image Bank and Metadata for Large-scale Research in Cerebrovascular Disease: Our Experience from the Stroke Image Bank Project Frontiers in ICT 2016: Paper 10.3389/fict.2016.00032. (2016) Folyóiratcikk /Szakcikk /Tudományos [16274394]
74. Kósa István , Vassányi István , Pintér Balázs , Dévényi Csaba Az otthoni monitorozás új európai tendenciái IME: INTERDISZCIPLINÁRIS MAGYAR EGÉSZSÉGÜGY / INFORMATIKA ÉS MENEDZSMENT AZ EGÉSZSÉGÜGYBEN 9:(4) pp. 43-46. (2010) Link(ek): Teljes dokumentum Folyóiratcikk /Szakcikk /Tudományos [1881780] [ Admin láttamozott ] Független idéző: 2 Összesen: 2 1.
Sebők Dávid, Szücs Veronika, Síkné Lányi Cecília Kinect-tel vezérelt stroke terápiás rendszer prototípusa In: XXIX. Neumann Kollokvium. Szeged, Magyarország: 2016.12.01-2016.12.02. Szeged: NJSzT, (2016.) , pp. 41-44 . ISBN: 978-963-306-514-3 (Neumann János Számítógéptudományi Társaság)
Egyéb konferenciaközlemény /Konferenciaközlemény /Tudományos [16249918] 2
Bán Attila A telekardiológia és a TTEKG megjelenése és szerepe az alapellátásban--háziorvosi interjúk tapasztalatai IME: INTERDISZCIPLINÁRIS MAGYAR EGÉSZSÉGÜGY / INFORMATIKA ÉS MENEDZSMENT AZ EGÉSZSÉGÜGYBEN (ISSN: 1588-6387) XVI: (4) pp. 41-44. (2017) Folyóiratcikk /Szakcikk /Tudományos [16605632]
75. Mák E , Pintér B , Gaál B , Vassányi I , Kozmann Gy , Németh I A Formal Domain Model for Dietary and Physical Activity Counseling In: Setchi R , Jordanov I , Howlett RJ , Jain LC (szerk.) Knowledge-Based and Intelligent Information and Engineering Systems: 14th International Conference, KES 2010 . 679 p. Konferencia helye, ideje: Cardiff , Egyesült Királyság / Wales , 2010.09.08 -2010.09.10. Berlin: Springer Verlag, 2010. pp. 607-616. ( Lecture Notes in Artificial Intelligence; 1. ) (ISBN:978-3-642-15386-0; 978-3-642-15387-7) Link(ek): WoS, Scopus, Egyéb URL Befoglaló mű link(ek): Teljes dokumentum Könyvrészlet /Konferenciaközlemény /Tudományos [1533067] [ Hitelesített ] Független idéző: 2 Összesen: 2 1
Zanamwe Ngonidzashe, Dube Kudakwashe, Thomson Jasmine S, Mtenzi Fredrick J, Hapanyengwi Gilford T Characterisation of Knowledge Incorporation into Solution Models for the Meal Planning Problem In: International Symposium on Foundations of Health Informatics Engineering and Systems. Berlin; Heidelberg: Springer, 2013. pp. 254-273. Link(ek): Scopus, Egyéb URL
Könyvrészlet /Könyvfejezet /Tudományos [16400832]
31
2
UYAR OKAN Preparing Diet List Suggestion with Fuzzy Expert System International Journal of Intelligent Systems and Applications in Engineering(IJISAE) 4: (Spec) pp. 58-62. (2016) Link(ek): Teljes dokumentum
Folyóiratcikk /Szakcikk /Tudományos [16400873]
76. Mák E , Pintér B , Gaál B , Vassányi I , Kozmann G , Németh I A formal domain model for dietary and physical activity counseling In: Setchi R , Jordanov I , Howlett RJ , Jain LC (szerk.) Knowledge-Based and Intelligent Information and Engineering Systems: 14th International Conference, KES 2010 . 679 p. Konferencia helye, ideje: Cardiff , Egyesült Királyság / Wales , 2010.09.08 -2010.09.10. Berlin: Springer Verlag, 2010. Paper formal domain. 10 p. ( Lecture Notes in Artificial Intelligence; 1. ) (ISBN:978-3-642-15386-0; 978-3-642-15387-7) Link(ek): DOI, Scopus Befoglaló mű link(ek): Teljes dokumentum Könyvrészlet /Konferenciaközlemény /Tudományos [1881821] [ Szerzői rekord ]
2009 77. Bognár Attila , Vassányi István , Végső Balázs , Dulai Tibor , Tarjányi Zsolt , Kozmann György , Kósa István Alpha: Otthoni távmonitorozás és döntéstámogatás IME: INTERDISZCIPLINÁRIS MAGYAR EGÉSZSÉGÜGY / INFORMATIKA ÉS MENEDZSMENT AZ EGÉSZSÉGÜGYBEN 8:(8) pp. 52-55. (2009) Link(ek): Teljes dokumentum Folyóiratcikk /Szakcikk /Tudományos [1881473] [ Admin láttamozott ] 78. Vassányi I , Gaál B , Fülöp K , Mák E , Kozmann G Personalized Dietary Counseling for Tele-care using Evolutionary Programming and Ontological Reasoning Med-e-Tel In: Jordanova M (szerk.) Global Telemedicine and eHealth Updates: Knowledge Resources, Vol. 2 . Konferencia helye, ideje: Luxembourg , Luxemburg , 2009.04 Luxembourg: Luxexpo, pp. 272-276. Egyéb konferenciaközlemény /Konferenciaközlemény /Tudományos [141055] [ Admin láttamozott ] Független idéző: 1 Összesen: 1 1.
Cseh Lajos Tamás, Kloó Norbert A Lavinia életmód-tükör szolgáltatás architekturális áttekintése In: Az e-Health kihívásai. A XXVI. Neumann Kollokvium kiadványa. Veszprém, Magyarország: 2013.11.22-2013.11.23. Veszprém: Pannon Egyetem, (2013.) , pp. 103-106 . ISBN: 978-615-5044-90-8 (Neumann János Számítógéptudományi Társaság)
Egyéb konferenciaközlemény /Konferenciaközlemény /Tudományos [16115404]
2008 79. Mák E , Gaál B , Vassányi I , Szabolcs I Igényfelmérés célzott betegek körében intelligens étrendtervező szoftverre Magyar Táplálkozástudományi Társaság 33.Vándorgyűlése , 2008.október 2-4. Pécs (2008) Link(ek): Egyéb URL Egyéb /Nem besorolt /Tudományos [1536915] [ Érvényesített ] 80. Mák E , Gaál B , Vassányi I , Karamánné Pakai A , Szabolcs I Egészségügyi szoftverek mesterséges intelligenciával - étrendtervező szoftver MAGYAR ORVOS 16:(11) pp. 36-38. (2008) Link(ek): MOB, Teljes dokumentum Folyóiratcikk /Rövid közlemény /Tudományos [1607597] [ Admin láttamozott ]
2007 81. Gaál B , Vassányi I , Kozmann G Application of artificial intelligence for weekly dietary menu planning In: Vaidya S , Yoshida Hiroyuki (szerk.) Advanced Computational Intelligence Paradigms in Healthcare . (65) Berlin: Berlin, Springer, 2007. pp. 27-48. ( Springer Studies in Computational Intelligence; 65. ) (ISBN:978-3-540-72374-5) Link(ek): DOI, Scopus Könyvrészlet /Konferenciaközlemény /Tudományos [1881822] [ Szerzői rekord ]
32
Független idéző: 7 Összesen: 7 1.
Snae C, Brückner M FOODS: A Food-Oriented Ontology-Driven System In: 2008 2nd IEEE International Conference on Digital Ecosystems and Technologies, IEEE-DEST 2008. Phitsanulok, 2008.02.26-2008.02.29. (2008.) , pp. 168-176 . Paper 4635195. Link(ek): DOI, Scopus
Egyéb konferenciaközlemény /Konferenciaközlemény [13723645] 2
Seljak BK Computer-based dietary menu planning: How to support it by complex knowledge? LECTURE NOTES IN ARTIFICIAL INTELLIGENCE (ISSN: 0302-9743) 6276 LNAI: (PART 1) pp. 587-596. (2010) Link(ek): DOI, Scopus
Folyóiratcikk /Tudományos [13723644] 3.
Joanna Ko ł odziejczyk, Ł ukasz Przyby ł ek Automatic dietary menu planning based on evolutionary algorithm In: Konferencja Polskiego Towarzystwa Badań Operacyjnych i Systemowych BOS2012. Varsó, Lengyelország: 2012.09.17-2012.09.19. (2012.) , pp. 1 .
Egyéb konferenciaközlemény /Konferenciaközlemény /Tudományos [16126564] 4
Cheng-Huang Lu Electronic device and recipe selecting method thereof 12 p. Szabadalmi szám/ügyiratszám: US 20130151550 A1. Benyújtás éve: 2011. Benyújtás száma: US 13/339,377. Benyújtás helye: Amerikai Egyesült Államok. Közzététel éve: 2013.
Oltalmi formák /USA szabadalom /Tudományos [15983738] 5
Richard Fox, Yuliya Bui An Artificial Intelligence Approach to Nutritional Meal Planning for Cancer Patients: Proceedings of the 4th Computer Science On-line Conference 2015 (CSOC2015), Vol 1: Artificial Intelligence Perspectives and Applications In: Radek Silhavy, Roman Senkerik, Zuzana Kominkova Oplatkova, Zdenka Prokopova, Petr Silhavy (szerk.) : Artificial Intelligence Perspectives and Applications. (347) Springer, 2015. (ISBN 978-3-319-18475-3) pp. 215-224. (Advances in Intelligent Systems and Computing) Link(ek): DOI
Könyvrészlet /Könyvfejezet /Tudományos [15983716] 6
Petra Oreskovic, Jasenka Gajdos Kljusuric, Zvonimir Satalic Computer-generated vegan menus: The importance of food composition database choice JOURNAL OF FOOD COMPOSITION AND ANALYSIS (ISSN: 0889-1575) 37: pp. 112-118. (2015) Link(ek): DOI
Folyóiratcikk /Szakcikk /Tudományos [15972398] 7.
Noppon Choosri, Sathita Anprasertphon Hospital dietary planning system using constraint programming In: Innovative Computing Technology (INTECH), 2015 Fifth International Conference on. Vigo, Spanyolország: 2015.05.20-2015.05.22. IEEE, (2015.) , pp. 17-22 . ISBN: 9781467375528 (IEEE) Link(ek): DOI
Egyéb konferenciaközlemény /Konferenciaközlemény /Tudományos [15983689]
82. Tarjan J , Vassanyi I , Voros B The value of common carotid artery intima-media thickness in evaluating atherosclerosis EUROPEAN HEART JOURNAL 28:(1) p. 113. (2007) Link(ek): WoS Folyóiratcikk /Szakcikk /Tudományos [3062126] [ Admin láttamozott ] 83. Vassányi István , Végső Balázs , Tarjányi Zsolt , Szakolczai Krisztina , Kozmann György , Kósa István Többparaméteres intelligens távmonitorozás IME: INTERDISZCIPLINÁRIS MAGYAR EGÉSZSÉGÜGY / INFORMATIKA ÉS MENEDZSMENT AZ EGÉSZSÉGÜGYBEN 6:(7) pp. 47-50. (2007) Link(ek): Teljes dokumentum Folyóiratcikk /Szakcikk /Tudományos [1881781] [ Admin láttamozott ]
2006 84. B Gaál , I Vassányi , Gy Kozmann , E Mák , I Szabolcs Expert System for Lifestyle and Nutrition Counseling - Életmód és táplálkozás-tanácsadó szakértői rendszer (2006) Egyéb /Nem besorolt /Tudományos [1536892] [ Szerzői rekord ] 85. Gaál B , Vassanyi I , Mák E , Kozmann Gy Életmód- és táplálkozás tanácsadó szakértői rendszer bemutatása 324. Tudományos Kollokvium az MTA Élelm. Tud. Komplex Bizottsága, a Központi Élelm.-tud. Kutatóint, és a Magyar Élelmezésipari Tud Egy. rendezvénye (2006) Egyéb /Nem besorolt /Tudományos [1536885] [ Admin láttamozott ]
33
86. S Palágyi , O Csirke , J Futó , J Hlavay , B Raucsik , A Szabó , I Vassányi Mining data from Roman sandstone quarries ACTA ARCHAEOLOGICA ACADEMIAE SCIENTIARUM HUNGARICAE 57: pp. 395-422. (2006) Folyóiratcikk /Szakcikk /Tudományos [1204474] [ Szerzői rekord ] 87. Vassányi István , Fogarassyné Vathy Ágnes , Tobak Tamás , Rovnyai János Adatbányászati alkalmazások az egészségügyben IME: INTERDISZCIPLINÁRIS MAGYAR EGÉSZSÉGÜGY / INFORMATIKA ÉS MENEDZSMENT AZ EGÉSZSÉGÜGYBEN 5:(5) pp. 49-53. (2006) Link(ek): Teljes dokumentum Folyóiratcikk /Szakcikk /Tudományos [1881782] [ Szerzői rekord ]
2005 88. Á Vathy-Fogarassy , G Balázs , T Tobak , I Vassányi Intelligent Data Analysis Center: A Client/Server Mining Model over the Internet In: Proceedings of 1st ADBIS Workshop on Data Mining and Knowledge Discovery: ADMKD'2005 . Konferencia helye, ideje: Tallinn , Észtország , 2005.09.15 -2005.09.16. pp. 57-65. Egyéb konferenciaközlemény /Konferenciaközlemény /Tudományos [1758616] [ Szerzői rekord ] 89. Fogarassyné Vathy Ágnes , Vassányi István Adatbányászati technológiák az egészségügyben IME: INTERDISZCIPLINÁRIS MAGYAR EGÉSZSÉGÜGY / INFORMATIKA ÉS MENEDZSMENT AZ EGÉSZSÉGÜGYBEN 4:(3) pp. 46-51. (2005) Link(ek): Teljes dokumentum Folyóiratcikk /Szakcikk /Tudományos [1881783] [ Szerzői rekord ] 90. Gaál B , Vassányi I , Kozmann G Automated planning of weekly dietary menus for personalized nutrition counselling In: Artificial Intelligence and Applications. Proceedings of the 23rd IASTED International Conference . pp. 300-305. Egyéb konferenciaközlemény /Konferenciaközlemény /Tudományos [140300] [ Admin láttamozott ] Független idéző: 1 Összesen: 1 1
Swati V Chande, Madhavi Sinha Genetic Algorithm: A Versatile Optimization Tool BVICAM’S International Journal of Information Technology (ISSN: 0973-5658) 2008: pp. 7-13. (2008) Folyóiratcikk /Szakcikk /Tudományos [15972428]
91. Gaál B , Vassányi I , Kozmann G A Novel Artificial Intelligence Method for Weekly Dietary Menu Planning METHODS OF INFORMATION IN MEDICINE 44: pp. 655-664. (2005) Link(ek): WoS, Scopus Folyóiratcikk /Szakcikk /Tudományos [140411] [ Érvényesített ] Független idéző: 24 Összesen: 24 1
Pennington JAT, Stumbo PJ, Murphy SP, Mcnutt SW, Eldridge AL, McCabe-Sellers BJ, Chenard CA Food composition data: The foundation of dietetic practice and research JOURNAL OF THE AMERICAN DIETETIC ASSOCIATION (ISSN: 0002-8223) 107: (12) pp. 2105-2113. (2007) Link(ek): DOI, PubMed, WoS, Scopus
Folyóiratcikk /Szakcikk [199968] 2
Samarghitean C, Vihinen M Medical expert systems CURRENT BIOINFORMATICS (ISSN: 1574-8936) 3: (1) pp. 56-65. (2008) Link(ek): DOI, WoS, Scopus
Folyóiratcikk /Szakcikk [199967] 3
Stephen L Smith, Stefano Cagnoni A Review of Medical Applications of Genetic and Evolutionary Computation In: Genetic and Evolutionary Computation: Medical Applications. Wiley, 2010. (ISBN 978-0-470-74813-8) pp. 17-43. Link(ek): DOI
Könyvrészlet /Könyvfejezet /Tudományos [15972450] 4.
Carlos A Coello Coello Fundamentals of Evolutionary Multi-Objective Optimization 23 p. http://delta.cs.cinvestav.mx/~ccoello/chapters/chapter-bogdan.pdf.gz. (2010.)
Egyéb /Kutatási jelentés (közzétett) /Tudományos [15948905]
34
5
Seljak BK Web-based eHealth applications with reference to food composition data EUROPEAN JOURNAL OF CLINICAL NUTRITION (ISSN: 0954-3007) 64: (Suppl. 3) pp. S121-S127. (2010) Link(ek): DOI, WoS, Scopus
Folyóiratcikk [12026368] 6.
Funabiki N, Taniguchi S, Matsushima Y, Nakanishi T A proposal of a menu planning algorithm for two-phase cooking by busy persons In: Proceedings of the International Conference on Complex, Intelligent and Software Intensive Systems, CISIS 2011. 2011.06.30-2011.07.02. Szöul: (2011.) , pp. 668-673 . Paper 5989119. ISBN: 9780769543734 Link(ek): DOI, Scopus
Egyéb konferenciaközlemény [12026373] 7.
Maya Sappelli An adaptive recipe recommendation system for people with Diabetes type 2 : Master’s Thesis in Artificial Intelligence 105 p. http://theses.ubn.ru.nl/bitstream/handle/123456789/190/Sappelli_MA_Thesis_2011.pdf?sequence=1. (2011.)
Egyéb /Diplomamunka, szakdolgozat, TDK dolgozat /Tudományos [15948900] 8.
Kovásznai G Developing an expert system for diet recommendation In: SACI 2011 - 6th IEEE International Symposium on Applied Computational Intelligence and Informatics, Proceedings. (1) 2011.05.19-2011.05.21. Temesvár: IEEE, (2011.) , pp. 505-509 . Paper 5873056. ISBN: 9781424491094 (IEEE) Link(ek): DOI, Scopus
Egyéb konferenciaközlemény /Konferenciaközlemény /Tudományos [12026372] 9.
Jarmo T Alander An Indexed Bibliography of Genetic Algorithms in Agriculture : Report Series No. 94-1-AGRO 67 p. http://lipas.uwasa.fi/~TAU/reports/report94-1/gaAGRObib.pdf. (2012.)
Egyéb /Kutatási jelentés (közzétett) /Tudományos [15948896] 10
J Chan Web-Based Decision Support System for School Meal Planning INTERNATIONAL JOURNAL OF INFORMATION SYSTEMS AND SOCIAL CHANGE (ISSN: 1941-868X) 3: (1) pp. 10-21. (2012) Link(ek): DOI
Folyóiratcikk [12917502] 11.
Cseh Lajos Tamás, Kloó Norbert A Lavinia életmód-tükör szolgáltatás architekturális áttekintése In: Az e-Health kihívásai. A XXVI. Neumann Kollokvium kiadványa. Veszprém, Magyarország: 2013.11.22-2013.11.23. Veszprém: Pannon Egyetem, (2013.) , pp. 103-106 . ISBN: 978-615-5044-90-8 (Neumann János Számítógéptudományi Társaság)
Egyéb konferenciaközlemény /Konferenciaközlemény /Tudományos [16115441] 12
Funabiki N, Matsushima Y, Nakanishi T, Watanabe K An extension of menu planning algorithm for two-phase homemade cooking In: 2013 IEEE 2nd Global Conference on Consumer Electronics, GCCE 2013. 2013. (ISBN 9781479908929) pp. 352-356. Paper 6664853. Link(ek): DOI, Scopus
Könyvrészlet /Konferenciaközlemény /Tudományos [14460720] 13
Pop CB, Chifu VR, Salomie I, Cozac A, Mesaros I Particle Swarm Optimization-based method for generating healthy lifestyle recommendations In: 2013 IEEE 9th International Conference on Intelligent Computer Communication and Processing, ICCP 2013. 2013. (ISBN 9781479914937) pp. 15-21. Paper 6646074. Link(ek): DOI, Scopus
Könyvrészlet /Konferenciaközlemény /Tudományos [14460719] 14.
Jarmo T Alander An Indexed Bibliography of Genetic Algorithms in Biosciences : Report Series No. 94-1-BIO 179 p. http://lipas.uwasa.fi/~TAU/reports/report94-1/gaBIObib.pdf. (2014.)
Egyéb /Kutatási jelentés (közzétett) /Tudományos [15948886] 15.
Jarmo T Alander An Indexed Bibliography of Genetic Algorithms in Medicine : Report Series No. 94-1-MEDICINE 147 p. http://lipas.uwasa.fi/~TAU/reports/report94-1/gaMEDICINEbib.pdf. (2014.)
Egyéb /Kutatási jelentés (közzétett) /Tudományos [15948892] 16
Dixon E, Condrasky MD, Corr A, Kemper K, Sharp J Application of a menu-planning template as a tool for promoting healthy preadolescent diets TOPICS IN CLINICAL NUTRITION (ISSN: 0883-5691) 29: (1) pp. 47-56. (2014) Link(ek): DOI, Scopus
Folyóiratcikk /Tudományos [14460718] 17.
Ngonidzashe Zanamwe, Kudakwashe Dube, Jasmine S Thomson, Fredrick J Mtenzi, Gilford T Hapanyengwi Characterisation of Knowledge Incorporation into Solution Models for the Meal Planning Problem In: Third International Symposium, FHIES 2013, Revised Selected Papers. (8315) Macau, Kína: 2013.08.21-2013.08.23. Springer, (2014.) , pp. 254-273 . Paper 10.1007/978-3-642-53956-5_17. ISBN: 978-3-642-53955-8 ISSN: 0302-9743 (Springer)
Egyéb konferenciaközlemény /Konferenciaközlemény /Tudományos [14508778] 18
Oscar Chávez - Bosquez, Jerusa Marchi, Pilar Pozos - Parra Félix Castro Espinoza (szerk.) Nutritional Menu Planning: A Hybrid Ap proach and Preliminary Tests Instituto Politécnico Nacional , Centro de Investigación en Computación, Mexico City, 2014. ( Research in Computing Science; 82. ) 11 p.
Könyv /Tanulmánykötet /Tudományos [15948917] 19
Cristina Bianca Pop, Viorica Rozina Chifu, Ioan Salomie, Cristian Prigoana, Tiberiu Boros, Dorin Moldovan Generating Healthy Menus for Older Adults Using a Hybrid Honey Bees Mating Optimization Approach Konferencia helye, ideje: Temesvár , Románia, 2015.09.21 -2015.09.24. Temesvár: IEEE, 2015. (IEEE) Link(ek): DOI, WoS
Egyéb konferenciakötet /Tudományos [15948940]
35
20
Viorica R Chifu, Ioan Salomie, Emil Şt Chifu, Cristina Bianca Pop, Dan Valea, Madalina Lupu, Marcel Antal Hybrid Invasive Weed Optimization Method for Generating Healthy Meals In: Valentina Emilia Balas, Lakhmi C Jain, Branko Kovačević (szerk.) : Soft Computing Applications. (365) Springer, 2015. (ISBN 978-3-319-18296-4) pp. 335-351. (Advances in Intelligent Systems and Computing) Link(ek): DOI
Könyvrészlet /Könyvfejezet /Tudományos [15948956] 21.
Jarmo T Alander An Indexed Bibliography of Genetic Algorithms in Geosciences, Astronomy, Aerospace Engineering, and Aerodynamics : Report Series No. 94-1-AERO 109 p. http://lipas.uwasa.fi/~TAU/reports/report94-1/gaAERObib.pdf. (2016.)
Egyéb /Kutatási jelentés (közzétett) /Tudományos [15948893] 22.
Juan Manuel Ramos Perez Multi-Objective Optimization Techniques applied to the Menu Planning Problem 69 p. (2016.)
Egyéb /Diplomamunka, szakdolgozat, TDK dolgozat /Tudományos [15948973] 23.
Viorica Rozina Chifu Razvan Mircea Bonta Emil stefan Chifu Ioan Salomie Dorin Moldovan Particle Swarm Optimization Based Method for Personalized Menu Recommendations In: International Conference on Advancements of Medicine and Health Care through Technology. Kolozsvár, Románia: 2016.10.12-2016.10.15. (2016.) , pp. 232-237 . (IFMBE) Link(ek): DOI
Egyéb konferenciaközlemény /Konferenciaközlemény /Tudományos [16605627] 24
Felix Mata, Miguel Torres-Ruiz, Roberto Zagal, Giovanni Guzmán, Marco Moreno, Rolando Quintero A cross-domain framework for designing healthcare mobile applications mining social networks to generate recommendations of training and nutrition planning TELEMATICS AND INFORMATICS (ISSN: 0736-5853) 2017: Paper 10.1016/j.tele.2017.04.005. (2017) Link(ek): DOI
Folyóiratcikk /Szakcikk /Tudományos [16514742]
92. Gaál B , Vassányi I , Kozmann Gy An Evolutionary Divide and Conquer Method for Long-Term Dietary Menu Planning LECTURE NOTES IN ARTIFICIAL INTELLIGENCE 3581: pp. 419-423. (2005) Link(ek): DOI Folyóiratcikk /Konferenciaközlemény /Tudományos [2442476] [ Admin láttamozott ] Független idéző: 1 Összesen: 1 1
Stephen L Smith, Stefano Cagnoni A Review of Medical Applications of Genetic and Evolutionary Computation In: S L Smith, S Cagnoni (szerk.) : Genetic and Evolutionary Computation: Medical Applications. Wiley, 2010. pp. 17-43. Link(ek): DOI
Könyvrészlet /Könyvfejezet /Tudományos [15983998]
93. Vassanyi I , Barcza J , Tobak T Making Two-Level Standards Work in Legacy Communication Frameworks In: Rolf Engelbrecht , Antoine Geissbuhler , Christian Lovis , George Mihalas (szerk.) Connecting Medical Informatics and Bio-Informatics - Proceedings of MIE2005 – The XIXth International Congress of the European Federation for Medical Informatics . Konferencia helye, ideje: Canton of Geneva , Svájc , 2005.08.28 -2005.09.05. Amsterdam: IOS Press, 2005. pp. 689-694. ( Studies in Health Technology and Informatics; 116. ) (ISBN:978-1-58603-549-5) Link(ek): WoS Könyvrészlet /Konferenciaközlemény /Tudományos [1881803] [ Szerzői rekord ] Független idéző: 1 Összesen: 1 1
Bell Raj Eapen ONTODerm - A domain ontology for dermatology DERMATOLOGY ONLINE JOURNAL (ISSN: 1087-2108) 14: (6) Paper 16. (2008) Folyóiratcikk /Szakcikk /Tudományos [15984012]
2004 94. Balazs G , Kozmann G , Vassanyi I Intelligent cardiac telemonitoring system In: IEEE (szerk.) Computers in Cardiology . 802 p. Konferencia helye, ideje: Chicago , Amerikai Egyesült Államok , 2004.09.19 -2004.09.22. Piscataway: IEEE Operations Center, 2004. pp. 745-748. ( COMPUTERS IN CARDIOLOGY; 31. ) (ISBN:0-7803-8927-1) Link(ek): WoS Könyvrészlet /Konferenciaközlemény /Tudományos [1881805] [ Szerzői rekord ] Független idéző: 3 Összesen: 3 1
Augustyniak Piotr, Izworski Andrzej
36
ECG interpretation task sharing rules in a distributed wearable monitoring system IASTED International Conference on Signal and Image Processing (ISSN: 1482-7921) 2006: pp. 302-307. (2006) Link(ek): WoS
Folyóiratcikk /Konferenciaközlemény /Tudományos [15683565] 2
Augustyniak Piotr Quality-driven continuous adaptiation of ECG interpretation in a distributed surveillance system LECTURE NOTES IN ARTIFICIAL INTELLIGENCE (ISSN: 0302-9743) 5097: (Zakopane) pp. 909-918. (2008) Link(ek): WoS
Folyóiratcikk /Konferenciaközlemény /Tudományos [15683564] 3
Augustyniak P Management of Non-uniform Data Transfer in Cardiac Monitoring Systems with Adaptive Interpretation COMPUTERS IN CARDIOLOGY (ISSN: 0276-6574) 37: (Belfast) pp. 49-52. (2010) Link(ek): WoS
Folyóiratcikk /Konferenciaközlemény /Tudományos [15683563]
95. Mariusz Duplaga , Jrysztof Zielinski (szerk.) International Workshop on Overcoming the barriers to e-health growth in Enlarged Europe Konferencia helye, ideje: Krakkó , Lengyelország , 2004.12.05 -2004.12.07. 2004. (ISBN:8391664945) Könyv /Konferenciakötet /Tudományos [3128609] [ Szerzői rekord ]
2003 96. Gaál B , Vassányi I , Kozmann G Étkezési javaslat automatikus generálása táplálkozási és életmód-tanácsadó rendszerhez IME: INTERDISZCIPLINÁRIS MAGYAR EGÉSZSÉGÜGY / INFORMATIKA ÉS MENEDZSMENT AZ EGÉSZSÉGÜGYBEN 2:(3) pp. 51-55. (2003) Folyóiratcikk /Szakcikk /Tudományos [139965] [ Admin láttamozott ]
2002 97. Kozmann G , Jokuthy A , Viranyi V , Vassanyi I Methodological studies related to cardiovascular risk assessment In: Gy Surján , R Engelbrecht , P McNair (szerk.) Health Data in the Information Society: Proceedings of MIE 2002 . 847 p. Amsterdam: IOS Press, 2002. pp. 635-638. ( Studies in health technology and informatics; v. 90. ) (ISBN:1586032798) Link(ek): WoS Könyvrészlet /Konferenciaközlemény /Tudományos [1881808] [ Szerzői rekord ] 17th International Conference of the European-Federation-of-Medical-Informatics (MIE 2002), AUG 25-29, 2002., BUDAPEST, HUNGARY
2001 98. Balázs Gábor , Jókuthy András , Vassányi István Informatikai alkalmazások az egészségügyben KÓRHÁZ 8:(5) pp. 43-48. (2001) Folyóiratcikk /Szakcikk /Tudományos [1881784] [ Szerzői rekord ] Független idéző: 1 Összesen: 1 1.
Szuchy Krisztina, Havasi Anett, Szakonyi Benedek, Tóth Noémi, Unger Vivien, Kósa István Tapasztalatok multimédiás betekoktatóanyag klinikai alkalmazásával In: XXIX. Neumann Kollokvium. Szeged, Magyarország: 2016.12.01-2016.12.02. Szeged: NJSzT, (2016.) , pp. 31-34 . ISBN: 978-963-306-514-3 (NJSzT)
Egyéb konferenciaközlemény /Konferenciaközlemény /Tudományos [16249906]
99. Vassanyi I , Mogor E , Szakolczai K , Tarnay K Mobile access to a cardiology database STUDIES IN HEALTH TECHNOLOGY AND INFORMATICS 84: p. 881. (2001) Link(ek): DOI, WoS, Scopus Folyóiratcikk /Absztrakt / Kivonat /Tudományos [2115753] [ Érvényesített ] Független idéző: 2 Összesen: 2 1
Patrik Eklund, Johan Karlsson, Annica Näslund Mobile Pharmacology In: Advances in Hybrid Information Technology. (4413) Springer, 2007. (ISBN 978-3-540-77367-2) pp. 522-533. (Lecture Notes in Computer Science) Link(ek): DOI
Könyvrészlet /Konferenciaközlemény /Tudományos [15983945]
37
2.
Syed Farid Syed Adnan, Shah Alam Selangor, Habibah Hashim Medical mobile to web services application In: Signal Processing & Its Applications, 2009. CSPA 2009. 5th International Colloquium on. Kuala Lumpur, Malajzia: 2009.03.06-2009.03.08. IEEE, (2009.) , pp. 124-129 . ISBN: 978-1-4244-4151-8 (IEEE) Link(ek): DOI
Egyéb konferenciaközlemény /Konferenciaközlemény /Tudományos [15983949]
2000 100. Medve Anna , Drozdik Béla , Mógor Emil , Vassányi István , Tarnay Katalin Kardiológiai adatbázis mobil elérésének formális leírása’ In: Kozmann Gy , Szakolczai K (szerk.) XXII. Neumann Kollokvium: A számítástechnika orvosi és biológiai alkalmazásai . Konferencia helye, ideje: Veszprém , Magyarország , 2000.11.09 -2000.11.10. Veszprém: Neumann János Számítógép-tudományi Társaság (NJSZT), 2000. pp. 120-127. Könyvrészlet /Konferenciaközlemény /Tudományos [2149391] [ Szerzői rekord ] 101. Vassanyi I Implementing fine grain processor arrays on field-programmable logic INTEGRATED COMPUTER-AIDED ENGINEERING 7:(1) pp. 53-66. (2000) Link(ek): WoS Folyóiratcikk /Szakcikk /Tudományos [1881810] [ Szerzői rekord ]
1998 102. Reiner W Hartenstein , Andres Keevallik G.Goos (szerk.) Field-Programmable Logic and Applications From FPGAs to Computing Paradigm: 8th International Workshop, FPL '98 Tallinn, Estonia, August 31–September 3, 1998 Proceedings Konferencia helye, ideje: Tallinn , Észtország , 1998.08.31 -1998.09.03. Berlin: Berlin, Springer, 1998. 198 p. ( Lecture Notes in Computer Science; 1482. ) (ISBN:978-3-540-64948-9) Könyv /Konferenciakötet /Tudományos [3100596] [ Szerzői rekord ] 103. I Vassanyi Implementing processor arrays on FPGAs In: Reiner W Hartenstein , Andres Keevallik G.Goos (szerk.) Field-Programmable Logic and Applications From FPGAs to Computing Paradigm: 8th International Workshop, FPL '98 Tallinn, Estonia, August 31–September 3, 1998 Proceedings . 198 p. Konferencia helye, ideje: Tallinn , Észtország , 1998.08.31 -1998.09.03. Berlin: Berlin, Springer, 1998. pp. 446-450. ( Lecture Notes in Computer Science; 1482. ) (ISBN:978-3-540-64948-9) Könyvrészlet /Konferenciaközlemény /Tudományos [3100600] [ Szerzői rekord ] Független idéző: 3 Összesen: 3 1.
E Fabinai D Lavenier Using knapsack technique to place linear arras of FPGA 18 p. IRISA, France. No. 1335. (2000.)
Egyéb /Kutatási jelentés (közzétett) /Tudományos [15983760] 2.
D Walsh, Piotr Dudek A compact FPGA implementation of a bit-serial SIMD cellular processor array In: 13th International Workshop on Cellular Nanoscale Networks and their Applications. Turin, Olaszország: 2012.08.29-2012.08.31. IEEE, (2012.) , pp. 1-6 . ISBN: 978-1-4673-0287-6 ISSN: 2165-0144 (IEEE) Link(ek): DOI
Egyéb konferenciaközlemény /Konferenciaközlemény /Tudományos [15983758] 3
J B Nade 1, R V Sarwadnya The Soft Core Processors: A Review INTERNATIONAL JOURNAL OF INNOVATIVE RESEARCH IN ELECTRICAL, ELECTRONICS, INSTRUMENTATION AND CONTROL ENGINEERIN (ISSN: 2321-5526) 3: (12) pp. 197-203. (2015) Link(ek): DOI
Folyóiratcikk /Szakcikk /Tudományos [15983772]
1997 104. E Czoboly (szerk.) 9th Hungarian-Korean Seminar on Integrated Circuits and Devices Konferencia helye, ideje: Budapest, Magyarország, 1997.06.24-1997.06.26. Budapest: 1997.
38
Egyéb konferenciakötet / /Tudományos [3128256] [ Szerzői rekord ] 105. I Erényi , I Vassányi , T Nemes , É Nikodémusz , Z Katona FPGA-based Automated Datapath Design In: EUROMICRO'97 . Konferencia helye, ideje: Budapest , Magyarország , 1997.09.01 -1997.09.05. pp. 664-668. Egyéb konferenciaközlemény /Konferenciaközlemény /Tudományos [1881825] [ Szerzői rekord ] 106. Erenyi I , Vassanyi I , Nemes T , Nikodemusz E , Katona Z FPGA-based automated datapath design In: Ferenc Vajda (szerk.) 23RD EUROMICRO CONFERENCE - NEW FRONTIERS OF INFORMATION TECHNOLOGY, PROCEEDINGS . Konferencia helye, ideje: Budapest , Magyarország , 1997.09.01 -1997.09.04. (IEEE) Budapest: IEEE, 1997. pp. 664-668. ( EUROMICRO CONFERENCE - PROCEEDINGS ) (ISBN:0-8186-8130-6) Link(ek): WoS Könyvrészlet /Konferenciaközlemény /Tudományos [1881812] [ Szerzői rekord ] 107. K Fazekas , I Erenyi (szerk.) Workshop on Digital Methodologies and Applications for Multimedia and Signal Processing Konferencia helye, ideje: Budapest , Magyarország , 1997.10.27 -1997.10.29. Panem, 1997. (ISBN:9635451830) Könyv /Konferenciakötet /Tudományos [3128123] [ Szerzői rekord ] 108. Vassányi I FPGAs and cellular algorithms: Two implementation examples JOURNAL OF SYSTEMS ARCHITECTURE 43:(1-5) pp. 23-26. (1997) Link(ek): DOI, WoS, Scopus Folyóiratcikk /Szakcikk /Tudományos [1397836] [ Szerzői rekord ]
1996 109. Balazs G , Drozdik B , Vassanyi I , Mogor E Medical information systems for cardiac risk assessment and remote monitoring In: Surjan G (szerk.) HEALTH DATA IN THE INFORMATION SOCIETY . Konferencia helye, ideje: Budapest , Magyarország , 1996.10.13 -1996.10.16. New York: IOS Press, 1996. pp. 809-814. ( STUDIES IN HEALTH TECHNOLOGY AND INFORMATICS; 90. ) (ISBN:1-58603-279-8) Link(ek): WoS Könyvrészlet /Konferenciaközlemény /Tudományos [1881809] [ Szerzői rekord ] 110. Erenyi I , Vassanyi I FPGA-based fine grain processor array design considerations In: Thanos Stouraitis (szerk.) ICECS 96 - PROCEEDINGS OF THE THIRD IEEE INTERNATIONAL CONFERENCE ON ELECTRONICS, CIRCUITS, AND SYSTEMS, VOLS 1 AND 2 . Konferencia helye, ideje: Rhodes , Görögország , 1996.10.13 -1996.10.16. Rhodes: Institute of Electrical and Electronics Engineers (IEEE), 1996. pp. 659-662. (ISBN:0-7803-3650-X) Link(ek): WoS Könyvrészlet /Konferenciaközlemény /Tudományos [1881811] [ Szerzői rekord ] Független idéző: 3 Összesen: 3 1
A DeHon Element placement method and apparatus 25 p. Szabadalmi szám/ügyiratszám: US7210112 B2. Benyújtás éve: 2002. Benyújtás száma: US 10/643,772. Közzététel éve: 2007.
Oltalmi formák /USA szabadalom /Tudományos [15983874]
39
2
A DeHon Method and apparatus for network with multilayer metalization 44 p. Szabadalmi szám/ügyiratszám: US7285487 B2. Benyújtás éve: 2003. Benyújtás száma: US 10/897,582. Benyújtás helye: Amerikai Egyesült Államok. Közzététel éve: 2007.
Oltalmi formák /USA szabadalom /Tudományos [15983872] 3
A DeHon Fast router and hardware-assisted fast routing method 40 p. Szabadalmi szám/ügyiratszám: US7342414 B2. Benyújtás éve: 2002. Benyújtás száma: US 10/356,710. Benyújtás helye: Amerikai Egyesült Államok. Közzététel éve: 2008.
Oltalmi formák /USA szabadalom /Tudományos [15983871]
1995 111. I Erenyi (szerk.) Proc. of the COMBIO´95 Workshop on Computational Modeling in Biosciences Konferencia helye, ideje: Kecskemét, Magyarország, 1995.08.01-1995.08.03. KFKI, 1995. KFKI-1995-14/M-H Egyéb konferenciakötet / /Tudományos [3128242] [ Szerzői rekord ] 112. László Zombory (szerk.) Technical conference on CAD methods in Electronic and Information Processing System Design and Its Education Konferencia helye, ideje: Budapest, Magyarország, 1995.06.08-1995.06.10. Budapest: Technical University of Budapest, 1995. Egyéb konferenciakötet / /Tudományos [3128178] [ Szerzői rekord ]
1993 113. István Erényi , István Vassányi Textile Quality Control: Challenge for Computerized Texture Analysis In: I Erenyi I Erenyi (szerk.) Workshop on Efficient Texture Analysis: Advanced Methods, Applications . Konferencia helye, ideje: Patras , Görögország , 1993.12.01 -1993.12.03. Budapest: KFKI, pp. 66-71. Egyéb konferenciaközlemény /Konferenciaközlemény /Tudományos [3127482] [ Szerzői rekord ] 114. Trón Tibor , Móczár Géza (szerk.) Végzős konferencia '93 Konferencia helye, ideje: Budapest, Magyarország, 1993.04.13. 1993. Egyéb konferenciakötet / /Tudományos [3128037] [ Szerzői rekord ] 115. Vassányi István Elosztott és konfigurálható adatbáziskezelő rendszerek In: Trón Tibor , Móczár Géza (szerk.) Végzős konferencia '93 . Konferencia helye, ideje: Budapest , Magyarország , 1993.04.13 pp. 145-150. Egyéb konferenciaközlemény /Konferenciaközlemény /Tudományos [3128108] [ Admin láttamozott ]
40
A KÉRELMEZŐ TANANYAGFORMÁLÓ KÉSZSÉGÉT IGAZOLÓ DOKUMENTUMOK
41
42
PANNON EGYETEM Tárgytematika Félév:
2016/17/1
Tárgynév:
Adatbáziskezelő rendszerek alkalmazása
Tárgykód:
VEMKSA5144A
Felelős szervezet neve:
Villamosmérnöki és Információs Rendszerek Tanszék
Felelős szervezet kódja:
MIVIR
Tárgyfelelős neve:
Dr. Vassányi István
Oktatás célja:
Tantárgy tartalma:
Számonkérési és értékelési rendszere:
Kötelező és ajánlott irodalom:
2016.09.05. 18:07:28
NEPTUN.NET Egységes Tanulmányi Rendszer
43
1
PANNON EGYETEM Tárgytematika Félév:
2015/16/2
Tárgynév:
Adatbáziskezelő rendszerek megvalósítása
Tárgykód:
VEMIVI5154A
Felelős szervezet neve:
Villamosmérnöki és Információs Rendszerek Tanszék
Felelős szervezet kódja:
MIVIR
Tárgyfelelős neve:
Dr. Vassányi István
Oktatás célja:
Tantárgy tartalma:
Számonkérési és értékelési rendszere:
2016.02.01. 10:43:18
NEPTUN.NET Egységes Tanulmányi Rendszer
44
1
PANNON EGYETEM Tárgytematika Félév:
2015/16/2
Tárgynév:
Adatbáziskezelő rendszerek megvalósítása
Tárgykód:
VEMIVI5154A
Felelős szervezet neve:
Villamosmérnöki és Információs Rendszerek Tanszék
Felelős szervezet kódja:
MIVIR
Tárgyfelelős neve:
Dr. Vassányi István
Számonkérési és értékelési rendszere:
2016.02.01. 10:43:18
NEPTUN.NET Egységes Tanulmányi Rendszer
45
2
PANNON EGYETEM Tárgytematika Félév:
2015/16/2
Tárgynév:
Adatbáziskezelő rendszerek megvalósítása
Tárgykód:
VEMIVI5154A
Felelős szervezet neve:
Villamosmérnöki és Információs Rendszerek Tanszék
Felelős szervezet kódja:
MIVIR
Tárgyfelelős neve:
Dr. Vassányi István
Számonkérési és értékelési rendszere:
2016.02.01. 10:43:18
NEPTUN.NET Egységes Tanulmányi Rendszer
46
3
PANNON EGYETEM Tárgytematika Félév:
2015/16/2
Tárgynév:
Adatbáziskezelő rendszerek megvalósítása
Tárgykód:
VEMIVI5154A
Felelős szervezet neve:
Villamosmérnöki és Információs Rendszerek Tanszék
Felelős szervezet kódja:
MIVIR
Tárgyfelelős neve:
Dr. Vassányi István
Számonkérési és értékelési rendszere:
Kötelező és ajánlott irodalom:
2016.02.01. 10:43:18
NEPTUN.NET Egységes Tanulmányi Rendszer
47
4
PANNON EGYETEM Tárgytematika Félév:
2016/17/2
Tárgynév:
Egészségügyi adatbázisok
Tárgykód:
VEMIVIM554E
Felelős szervezet neve:
Villamosmérnöki és Információs Rendszerek Tanszék
Felelős szervezet kódja:
MIVIR
Tárgyfelelős neve:
Dr. Vassányi István
Oktatás célja:
Tantárgy tartalma:
Számonkérési és értékelési rendszere:
Kötelező és ajánlott irodalom:
2017.02.07. 13:40:25
NEPTUN.NET Egységes Tanulmányi Rendszer
48
1
PANNON EGYETEM Tárgytematika Félév:
2016/17/1
Tárgynév:
Információ és hírközléselmélet
Tárgykód:
VEMIIR3112I
Felelős szervezet neve:
Villamosmérnöki és Információs Rendszerek Tanszék
Felelős szervezet kódja:
MIVIR
Tárgyfelelős neve:
Dr. Vassányi István
Oktatás célja:
Tantárgy tartalma:
Számonkérési és értékelési rendszere:
2016.09.05. 17:57:05
NEPTUN.NET Egységes Tanulmányi Rendszer
49
1
PANNON EGYETEM Tárgytematika Félév:
2016/17/1
Tárgynév:
Információ és hírközléselmélet
Tárgykód:
VEMIIR3112I
Felelős szervezet neve:
Villamosmérnöki és Információs Rendszerek Tanszék
Felelős szervezet kódja:
MIVIR
Tárgyfelelős neve:
Dr. Vassányi István
Számonkérési és értékelési rendszere:
2016.09.05. 17:57:05
NEPTUN.NET Egységes Tanulmányi Rendszer
50
2
PANNON EGYETEM Tárgytematika Félév:
2016/17/1
Tárgynév:
Információ és hírközléselmélet
Tárgykód:
VEMIIR3112I
Felelős szervezet neve:
Villamosmérnöki és Információs Rendszerek Tanszék
Felelős szervezet kódja:
MIVIR
Tárgyfelelős neve:
Dr. Vassányi István
Számonkérési és értékelési rendszere:
2016.09.05. 17:57:05
NEPTUN.NET Egységes Tanulmányi Rendszer
51
3
PANNON EGYETEM Tárgytematika Félév:
2016/17/1
Tárgynév:
Információ és hírközléselmélet
Tárgykód:
VEMIIR3112I
Felelős szervezet neve:
Villamosmérnöki és Információs Rendszerek Tanszék
Felelős szervezet kódja:
MIVIR
Tárgyfelelős neve:
Dr. Vassányi István
Számonkérési és értékelési rendszere:
Kötelező és ajánlott irodalom:
2016.09.05. 17:57:05
NEPTUN.NET Egységes Tanulmányi Rendszer
52
4
PANNON EGYETEM Tárgytematika Félév:
2016/17/2
Tárgynév:
Információs rendszerek biztonságtechnikája
Tárgykód:
VEMISAM144B
Felelős szervezet neve:
Rendszer- és Számítástudományi Tanszék
Felelős szervezet kódja:
MISA
Tárgyfelelős neve:
Dr. Süle Zoltán
Oktatás célja:
Tantárgy tartalma:
Számonkérési és értékelési rendszere:
2017.02.20. 8:17:53
NEPTUN.NET Egységes Tanulmányi Rendszer
53
1
PANNON EGYETEM Tárgytematika Félév:
2016/17/2
Tárgynév:
Információs rendszerek biztonságtechnikája
Tárgykód:
VEMISAM144B
Felelős szervezet neve:
Rendszer- és Számítástudományi Tanszék
Felelős szervezet kódja:
MISA
Tárgyfelelős neve:
Dr. Süle Zoltán
Számonkérési és értékelési rendszere:
Kötelező és ajánlott irodalom:
2017.02.20. 8:17:53
NEPTUN.NET Egységes Tanulmányi Rendszer
54
2
PANNON EGYETEM Tárgytematika Félév:
2016/17/2
Tárgynév:
Információs rendszerek biztonságtechnikája
Tárgykód:
VEMISAM144B
Felelős szervezet neve:
Rendszer- és Számítástudományi Tanszék
Felelős szervezet kódja:
MISA
Tárgyfelelős neve:
Dr. Süle Zoltán
Kötelező és ajánlott irodalom:
2017.02.20. 8:17:53
NEPTUN.NET Egységes Tanulmányi Rendszer
55
3
Tárgykód VEMISAM144B VEMISAM144B VEMIVI1112D VEMIVI5154A VEMIVI5154A
Tárgy név Információs rendszerek biztonságtechnikája Információs rendszerek biztonságtechnikája Digitális technika I. Adatbáziskezelő rendszerek megvalósítása Adatbáziskezelő rendszerek megvalósítása
Tárgy kredit 5
Félév
Kurzus Kurzus típus kód
Létszám
2011/12/2 03
Elmélet/ -/ - 12/0/999
K:12:00-14:00(I.1.I122);
5
2011/12/2 04
Gyakorlat/ 12/0/999 /-
K:14:00-16:00(I.1.I122);
2
2011/12/2 03
Elmélet/ -/ - 43/0/300
CS:14:00-16:00(A.AL.A01);
4
2011/12/2 05
4
2011/12/2 06
11/0/50(min.& Elmélet/ -/ K:08:00-10:00(I.fsz.I2); nbsp;10) 11/0/50(min.& Labor/ -/ K:10:00-12:00(I.fsz.I2); nbsp;10) P:08:00-11:15(B.2.B203); Elmélet/ -/ - 17/0/999 P:15:15-18:30(B.2.B203);
VEMIIR3112I
Információ és hírközléselmélet 2
2012/13/1 06L
VEMIIR3112I
Információ és hírközléselmélet 2
2012/13/1 06
VEMIVI1112D
Digitális technika I.
2
2012/13/1 1213_I Elmélet/ -/ - 1/0/300
VEMIVI2112D
Digitális technika II.
2
2012/13/1 09
VEMKIR3114H VEMKSA5144A VEMKSA5144A VEMKSA5144A VEMKSA5144A VEMIIR3112I
Információ és hírközléselmélet 4 Adatbáziskezelő rendszerek alkalmazása Adatbáziskezelő rendszerek alkalmazása Adatbáziskezelő rendszerek alkalmazása Adatbáziskezelő rendszerek alkalmazása
Órarend inf.
Elmélet/ -/ - 50/0/999
Elmélet/ -/ - 37/0/200
K:08:00-10:00(I.fsz.I1);
CS:10:00-12:00(I.fsz.I1);
2012/13/1 1213_I Elmélet/ -/ - 1/0/999
Dr. Vassányi István
4
2012/13/1 70L
Elmélet/ -/ - 19/0/60
P:08:00-11:15(I.1.I103); P:15:15-18:30(I.1.I103);
Dr. Vassányi István
K:12:00-14:00(I.fsz.I2);
Dr. Vassányi István
P:08:00-11:15(I.1.I103); P:11:30-15:00(I.1.I103);
Dr. Vassányi István
4 4
2012/13/1 71 2012/13/1 71L 2012/13/2 06ER
Gyakorlat/ 44/0/60 /Gyakorlat/ 19/0/60 /Elmélet/ -/ - 2/0/2
Dr. Vassányi István
Elmélet/ -/ - 12/0/999
K:12:00-14:00(I.1.I104);
Gyakorlat/ 12/0/999 /Gyakorlat/ 12/0/999 /-
SZO:13:30-18:30(I.1.I104); SZO:17:00-18:30(I.1.I104);
Dr. Juhász Zoltán, Dr. Vassányi István, Sikné dr. Lányi Cecilia Dr. Juhász Zoltán, Sikné dr. Lányi Cecilia, Dr. Vassányi István Dávid Ákos, Smidla József, Dr. Süle Zoltán,... Dávid Ákos, Smidla József, Dr. Süle Zoltán,... Dávid Ákos, Smidla József, Dr. Süle Zoltán,... Dávid Ákos, Smidla József, Dr. Süle Zoltán,...
CS:14:00-16:00(I.2.I201);
Dr. Vassányi István
H:14:00-16:00(I.fsz.PC0);
VEMIIR3354W
Web-alapú rendszerek fejlesztése
4
2012/13/2 21
Labor/ -/ -
H:16:00-18:00(I.fsz.PC0);
5
2012/13/2 08
Elmélet/ -/ - 12/0/999
SZO:13:30-18:30(I.1.I104); SZO:15:15-18:30(I.1.I104);
5
2012/13/2 06
VEMIVI5154A VEMIVI5154A VEMIVIM522A VEMKSA5144A VEMKSA5144A
Digitális technika I. Adatbáziskezelő rendszerek megvalósítása Adatbáziskezelő rendszerek megvalósítása Egészségügyi adatszabványok Adatbáziskezelő rendszerek alkalmazása Adatbáziskezelő rendszerek alkalmazása
Dr. Vassányi István
K:10:00-12:00(I.fsz.I2);
Elmélet/ -/ - 47/0/50
VEMIVI1112D
Dr. Vassányi István
Elmélet/ -/ - 44/0/60
2012/13/2 20
VEMISAM144B
Dr. Vassányi István
2012/13/1 70
4
VEMISAM144B
Dr. Vassányi István
Dr. Vassányi István
Web-alapú rendszerek fejlesztése
VEMISAM144B
Dr. Vassányi István
4
Információ és hírközléselmélet 2
Információs rendszerek biztonságtechnikája Információs rendszerek biztonságtechnikája Információs rendszerek biztonságtechnikája Információs rendszerek biztonságtechnikája
Dr. Vassányi István
Dr. Vassányi István
VEMIIR3354W
VEMISAM144B
Oktatók Smidla József, Dr. Süle Zoltán, Dr. Vassányi István Smidla József, Dr. Süle Zoltán, Dr. Vassányi István
5
2012/13/2 07
47/0/50
K:14:00-16:00(I.1.I104);
5
2012/13/2 09
2
2012/13/2 04
Elmélet/ -/ - 29/0/300
4
2012/13/2 07
Elmélet/ -/ -
4
2012/13/2 08
Labor/ -/ -
3
2012/13/2 01
Gyakorlat/ 16/0/25 /-
Dr. Vassányi István
4
2012/13/2 70ER
Elmélet/ -/ - 2/0/2
Dr. Vassányi István
2012/13/2 71ER
Gyakorlat/ 2/0/2 /-
4
21/0/50(min.& H:08:00-10:00(I.1.I104); nbsp;10) 21/0/50(min.& H:10:00-12:00(I.1.I104); nbsp;10)
VEMIIR3112I
Információ és hírközléselmélet 2
2013/14/1 07
Elmélet/ -/ - 63/0/999
VEMIIR3112I
Információ és hírközléselmélet 2
2013/14/1 07L
Elmélet/ -/ - 18/0/999
VEMIVI2112D
Digitális technika II.
2
2013/14/1 10
Elmélet/ -/ - 27/0/200
VEMKSA5144A
Adatbáziskezelő rendszerek alkalmazása
4
2013/14/1 72
Elmélet/ -/ - 31/0/60
VEMKSA5144A
Adatbáziskezelő rendszerek alkalmazása
4
2013/14/1 72L
Elmélet/ -/ - 26/0/60
VEMKSA5144A
Adatbáziskezelő rendszerek alkalmazása
4
2013/14/1 73L
Gyakorlat/ 26/0/60 /-
56
Dr. Vassányi István Dr. Vassányi István
Dr. Vassányi István SZE:08:00-10:00(I.2.I201); SZE:08:00-10:00(I.2.I201); SZE:08:00-10:00(I.2.I201); SZE:08:00-10:00(... P:08:00-13:00(B.2.B210); P:11:30-15:00(B.2.B210); P:11:30-13:00(B.2.B210); P:11:30-15:00(B.2.B210); K:08:00-10:00(B.2.B202); K:08:00-10:00(B.2.B202); K:08:00-10:00(B.2.B202); K:08:00-10:00(B.2.B202... CS:12:00-14:00(I.fsz.PC0); CS:12:00-14:00(I.fsz.PC0); CS:12:00-14:00(I.fsz.PC0); CS:12:00-14:00(I... P:08:00-13:00(I.1.I104); P:13:30-18:30(I.1.I104); P:15:15-18:30(I.1.I104); P:15:15-18:30(I.1.I104); P:15:15-18:30(I.1.I104);
Dr. Vassányi István
Dr. Vassányi István
Dr. Vassányi István
Dr. Vassányi István
Dr. Vassányi István Dr. Vassányi István
VEMKSA5144A
Adatbáziskezelő rendszerek alkalmazása
4
2013/14/1 73
Gyakorlat/ 31/0/60 /-
CS:14:00-16:00(I.fsz.PC0); CS:14:00-16:00(I.fsz.PC0); CS:14:00-16:00(I.fsz.PC0); CS:14:00-16:00(I...
Dr. Vassányi István
VEMISAM144B
Információs rendszerek biztonságtechnikája
5
2013/14/2 10
Elmélet/ -/ - 19/0/999
K:12:00-14:00(I.1.I104);
Dr. Dávid Ákos, Dr. Süle Zoltán, Dr. Vassányi István,...
VEMISAM144B
Információs rendszerek biztonságtechnikája
5
2013/14/2 11
Gyakorlat/ 19/0/999 /-
K:14:00-16:00(I.1.I104);
Dr. Dávid Ákos, Dr. Süle Zoltán, Dr. Vassányi István,...
VEMIVI1112D
Digitális technika I.
2
2013/14/2 05
Elmélet/ -/ - 26/0/300
SZE:12:00-14:00(B.1.B101);
Dr. Vassányi István
H:14:00-16:00(I.1.I104);
Dr. Vassányi István
P:10:00-12:00(I.fsz.PC0);
Dr. Vassányi István
H:16:00-18:00(I.1.I104);
Dr. Vassányi István
P:12:00-14:00(I.fsz.PC0);
Dr. Vassányi István
VEMIVI5154A VEMIVI5154A VEMIVI5154A VEMIVI5154A VEMKIR3114H
Adatbáziskezelő rendszerek megvalósítása Adatbáziskezelő rendszerek megvalósítása Adatbáziskezelő rendszerek megvalósítása Adatbáziskezelő rendszerek megvalósítása
4 4 4 4
Információ és hírközléselmélet 4
16/0/50(min.& 2013/14/2 09 Elmélet/ -/ nbsp;10) 4/0/50(min.&n 2013/14/2 09_BR Elmélet/ -/ bsp;3) 16/0/50(min.& 2013/14/2 10 Labor/ -/ nbsp;10) 4/0/50(min.&n 2013/14/2 10_BR Labor/ -/ bsp;3) 2013/14/2 14_ER Elmélet/ -/ - 2/0/2
Dr. Vassányi István
VEMIIR3112I
Információ és hírközléselmélet 2
2014/15/1 08L
Elmélet/ -/ - 17/0/999
P:08:00-11:15(B.2.B205); P:08:00-13:00(B.2.B205); P:08:00-13:00(B.2.B205);
VEMIIR3112I
Információ és hírközléselmélet 2
2014/15/1 08
Elmélet/ -/ - 39/0/999
K:10:00-12:00(I.fsz.I1);
Dr. Vassányi István
VEMIVI2112D
Digitális technika II.
2
2014/15/1 11
Elmélet/ -/ - 33/0/200
K:08:00-10:00(I.fsz.I1);
Dr. Vassányi István
VEMKSA5144A
Adatbáziskezelő rendszerek alkalmazása
4
2014/15/1 74
Elmélet/ -/ - 45/0/60
CS:08:00-10:00(I.fsz.PC0);
Dr. Vassányi István
Elmélet/ -/ - 17/0/60
P:15:15-18:30(I.1.I103); P:15:15-18:30(I.1.I103); P:15:15-18:30(I.1.I103);
Dr. Vassányi István
VEMKSA5144A VEMKSA5144A VEMKSA5144A
Adatbáziskezelő rendszerek alkalmazása Adatbáziskezelő rendszerek alkalmazása Adatbáziskezelő rendszerek alkalmazása
4
2014/15/1 74L
4
2014/15/1 75
4
2014/15/1 75L
Dr. Vassányi István
Gyakorlat/ 45/0/60 /Gyakorlat/ 17/0/60 /-
CS:10:00-12:00(I.fsz.PC0);
Dr. Vassányi István
P:08:00-13:00(I.1.I103); P:13:30-18:30(I.1.I103);
Dr. Vassányi István
VEMISAM144B
Információs rendszerek biztonságtechnikája
5
2014/15/2 14
Elmélet/ -/ - 18/0/999
CS:11:00-13:00(I.1.I104);
Dr. Dávid Ákos, Dr. Süle Zoltán, Dr. Vassányi István,...
VEMISAM144B
Információs rendszerek biztonságtechnikája
5
2014/15/2 15
Gyakorlat/ 18/0/999 /-
CS:13:00-15:00(I.1.I104);
Dr. Dávid Ákos, Dr. Süle Zoltán, Dr. Vassányi István,...
VEMIVI1112D
Digitális technika I.
2
2014/15/2 06
Elmélet/ -/ - 62/0/100
SZE:10:00-12:00(I.fsz.I1);
Dr. Vassányi István
8/0/15(min.&n Elmélet/ -/ SZE:16:00-18:00(I.fsz.I3); bsp;5) 8/0/15(min.&n Labor/ -/ SZE:18:00-20:00(I.fsz.I3); bsp;5)
Dr. Vassányi István, Dr. Kósa István Dr. Vassányi István, Dr. Kósa István
VEMIVIM554E
Egészségügyi adatbázisok
5
2014/15/2 03
VEMIVIM554E
Egészségügyi adatbázisok
5
2014/15/2 04
VEMIIR3112I
Információ és hírközléselmélet 2
2015/16/1 09_L
Elmélet/ -/ - 3/0/999
P:08:00-11:15(I.4.414);
Dr. Vassányi István
VEMIIR3112I
Információ és hírközléselmélet 2
2015/16/1 09
Elmélet/ -/ - 20/0/999
K:10:00-12:00(I.fsz.I1);
Dr. Vassányi István
VEMIIR3112I
Információ és hírközléselmélet 2
2015/16/1 09_ER Elmélet/ -/ - 0/0/999
P:08:00-12:00(I.4.414);
Dr. Vassányi István
VEMIVI2112D
Digitális technika II.
2
2015/16/1 12
Elmélet/ -/ - 22/0/200
K:12:00-14:00(I.fsz.I1);
Dr. Vassányi István
VEMKSA5144A
Adatbáziskezelő rendszerek alkalmazása
4
2015/16/1 76
Elmélet/ -/ - 56/0/60
SZE:08:00-10:00(I.fsz.PC0);
Dr. Vassányi István
Elmélet/ -/ - 8/0/60
P:08:00-13:00(I.3.317); P:15:15-18:30(I.3.318 (CISCO Dr. Vassányi István labor));
VEMKSA5144A VEMKSA5144A VEMKSA5144A VEMISAM144B VEMISAM144B VEMIVI1112D VEMIVI5154A VEMIVI5154A
Adatbáziskezelő rendszerek alkalmazása Adatbáziskezelő rendszerek alkalmazása Adatbáziskezelő rendszerek alkalmazása Információs rendszerek biztonságtechnikája Információs rendszerek biztonságtechnikája Digitális technika I. Adatbáziskezelő rendszerek megvalósítása Adatbáziskezelő rendszerek megvalósítása
4
2015/16/1 76_L
4
2015/16/1 77
4
2015/16/1 77_L
5
Gyakorlat/ 56/0/60 /Gyakorlat/ 8/0/60 /-
SZE:10:00-12:00(I.fsz.PC0); P:08:00-11:15(I.3.317);
Dr. Vassányi István Dr. Vassányi István Rosta Imre, Dr. Dávid Ákos, Dr. Vassányi István Dr. Dávid Ákos, Rosta Imre, Dr. Vassányi István
2015/16/2 18
Elmélet/ -/ - 16/0/999
H:08:00-10:00(I.1.I104);
5
2015/16/2 19
Gyakorlat/ 16/0/999 /-
H:10:00-12:00(I.1.I104);
2
2015/16/2 07
Elmélet/ -/ - 56/0/100
SZE:10:00-12:00(I.fsz.I1);
Dr. Vassányi István
4
2015/16/2 11
SZE:16:00-18:00(I.fsz.I3);
Dr. Vassányi István
4
2015/16/2 12
VEMIVIM554E
Egészségügyi adatbázisok
5
2015/16/2 05
VEMIVIM554E
Egészségügyi adatbázisok
5
2015/16/2 06
13/0/50(min.& nbsp;10) 13/0/50(min.& Labor/ -/ nbsp;10) 1/0/15(min.&n Elmélet/ -/ bsp;1) 1/0/15(min.&n Labor/ -/ bsp;1) Elmélet/ -/ -
57
SZE:18:00-20:00(I.fsz.I3); K:12:00-14:00(I.1.I104); K:14:00-16:00(I.1.I104);
Dr. Vassányi István Dr. Vassányi István, Dr. Kósa István Dr. Vassányi István, Dr. Kósa István
VEMIIR3112I
Információ és hírközléselmélet 2
2016/17/1 10_L
Elmélet/ -/ - 10/0/999
P:08:00-11:15(I.4.414); P:08:00-11:15(I.4.414); P:15:15-18:30(I.4.414); P:15:15-18:30(I.fsz.I1);
Dr. Vassányi István
VEMIIR3112I
Információ és hírközléselmélet 2
2016/17/1 10
Elmélet/ -/ - 16/0/999
K:10:00-12:00(I.4.416);
Dr. Vassányi István
VEMIVI2112D
Digitális technika II.
2016/17/1 13
Elmélet/ -/ - 18/0/200
K:14:00-16:00(I.4.415);
Dr. Vassányi István
Elmélet/ -/ - 9/0/60
P:08:00-11:15(I.4.417); P:11:30-13:00(I.4.417); P:13:30-18:30(I.fsz.I3);
Dr. Vassányi István
VEMKSA5144A VEMKSA5144A VEMKSA5144A VEMKSA5144A
Adatbáziskezelő rendszerek alkalmazása Adatbáziskezelő rendszerek alkalmazása Adatbáziskezelő rendszerek alkalmazása Adatbáziskezelő rendszerek alkalmazása
2 4
2016/17/1 78_L
4
2016/17/1 78
Elmélet/ -/ - 47/0/50
SZE:10:00-12:00(I.fsz.PC0);
Dr. Vassányi István
2016/17/1 79
Gyakorlat/ 47/0/50 /-
SZE:08:00-10:00(I.fsz.PC0);
Dr. Vassányi István
Gyakorlat/ 9/0/60 /-
P:13:30-18:30(I.4.417); P:13:30-15:00(I.4.417); P:15:15-18:30(I.4.417);
Dr. Vassányi István
4 4
2016/17/1 79_L
58
59
60
NYILATKOZAT
Alulírott Vassányi István (egyetemi docens, PE MIK, Villamosmérnöki és Információs Rendszerek Tanszék) nyilatkozom, hogy nincs folyamatban semmilyen tudományterületen habilitációs eljárásom, és két éven belül nem utasították el habilitációs kérelmemet.
Veszprém, 2017.
Dr. Vassányi István egyetemi docens
61
62
63
ÖSSZESEN:
publikáció címe
18.34
év
típus
szerzők száma SCI IF
Elosztott és konfigurálható adatbáziskezelő rendszerek 1993 magyar_konf_cikk 1 Textile Quality Control: Challenge for Computerized Texture Analysis 1993 angol_konf_cikk 2 1994 Strategies for Mapping Parallel Algorithms Onto Array Processors 1994 angol_konf_cikk 2 Mapping image processing algorithms onto a configurable 2-D processor array 1994 angol_konf_cikk 1 1995 An FPGA-based Cellular Array for Binary Morphological Operations 1995 angol_konf_cikk 1 Implementation Considerations for Cellular Image Processing Algorithms on FPGAs 1995 angol_konf_cikk 1 Implementation of Cellular Image Processing Algorithms on FPGAs 1995 angol_konf_cikk 1 1996 Implementation of Processor Cells for Array Algorithms on FPGAs 1996 angol_konf_cikk 2 FPGA-based fine grain processor array design considerations 1996 angol_konf_cikk 2 1997 FPGA-based Reconfigurable Soft Processing Arrays 1997 angol_konf_cikk 2 Structural concepts for automated datapath design 1997 angol_konf_cikk 5 FPGA-based Automated Datapath Design 1997 angol_konf_cikk 5 FPGAs and cellular algorithms: Two implementation examples 1997 if_cikk 1 1998 Implementing processor arrays on FPGAs 1998 if_cikk 1 2000 Kardiológiai adatbázis mobil elérése 2000 magyar_konf_cikk 4 Kardiológiai adatbázis mobil elérésének formális leírása 2000 magyar_konf_cikk 5 Implementing fine grain processor arrays on field-programmable logic 2000 if_cikk 1 2001 Mobile access to a cardiology database 2001 abstract 4 Informatikai alkalmazások az egészségügyben 2001 mo_f_cikk 3 2002 Medical Information Systems for Cardiac Risk Assessment and Remote Monitoring 2002 angol_konf_cikk 4 Methodological studies related to cardiovascular risk assessment 2002 magyar_konf_cikk 4 2003 Automated planning of Weekly Menus for Personalized Cardiovascular Risk Counselling 2003 angol_konf_cikk 3 Információs referencia-modellek az egészségügyben 2003 mo_f_cikk 3 Étkezési javaslat automatikus generálása táplálkozási és életmód-tanácsadó rendszerhez 2003 mo_f_cikk 3 2004 Implementation Aspects os Reference Model Based Communication Frameworks 2004 angol_konf_cikk 2 Intelligent cardiac telemonitoring system 2004 angol_konf_cikk 3 2005 Római kori víznyerő helyek kutatásának tapasztalatai Veszprém megyében 2005 magyar_konf_cikk 5 A Novel Artificial Intelligence Method for Weekly Dietary Menu Planning 2005 if_cikk 3 Making Two-Level Standards Work in Legacy Communication Frameworks 2005 if_cikk 3 Adatbányászati technológiák az egészségügyben 2005 angol_konf_cikk 3 Intelligent Data Analysis Center: A Client/Server Mining Model over the Internet 2005 mo_f_cikk 2 An Evolutionary Divide and Conquer Method for Long-Term Dietary Menu Planning 2005 angol_konf_cikk 4 Automated planning of weekly dietary menus for personalized nutrition counselling 2005 angol_konf_cikk 3 2006 Adatbányászati alkalmazások az egészségügyben 2006 mo_f_cikk 4 Mining data from Roman sandstone quarries 2006 mo_f_cikk 7 2007 Application of artificial intelligence for weekly dietary menu planning 2007 könyvrészlet 3 Többparaméteres intelligens távmonitorozás 2007 mo_f_cikk 6 2008 Egészségügyi szoftverek mesterséges intelligenciával - étrendtervező szoftver 2008 mo_f_cikk 5 Az arteria carotis communis intima-media vastagságának értéke az ateroszklerózis megítélésében 2008 mo_f_cikk 3 Mapping Clinical Databases to the Neuroweb Ontology: Lessons Learned 2008 angol_konf_cikk 3 2009 Evolutionary Algorithms for Cardiovascular Decision Support 2009 angol_konf_cikk 6 Alpha: Otthoni távmonitorozás és döntéstámogatás 2009 mo_f_cikk 7 Personalized Dietary Counseling for Tele-care using Evolutionary Programming and Ontological 2009 angol_konf_cikk Reasoning 5 2010 Alpha System: a multi-parameter remote monitoring system to cover the requirements 2010 of aangol_konf_cikk polymorbid aging population6 A formal domain model for dietary and physical activity counseling 2010 angol_konf_cikk 6 Az otthoni monitorozás új európai tendenciái 2010 mo_f_cikk 4 An ontological modeling approach to cerebrovascular disease studies: the NEUROWEB2010 case.if_cikk 10 2011 Tele-monitoring for neurological patients: lessons learned 2011 angol_konf_cikk 8 Applications of medical intelligence in remote monitoring 2011 angol_konf_cikk 10 Stress ECG utilization in the evaluation of patients with chest pain: The real practice in2011 Hungary if_cikk with 10 million inhabitants5 Orvosi intelligencia alkalmazásai a távdiagnosztikában 2011 mo_f_cikk 6 Táplálkozás-tanácsadó szakértői rendszer 2011 mo_f_cikk 5 Applications of Medical Intelligence in Remote Monitoring 2011 külf_f_cikk 10 2012 MenuGene: A Comprehensive Expert System for Dietary and Lifestyle Counseling and 2012 Tracking angol_konf_cikk 4 GPU based parallel genetic algorithm library 2012 angol_konf_cikk 5
64
oldalszám
12.05
65
MTA Q-szám
MTA I-szám (hiv.)
0.10 0.10 0.10 0.20 0.20 0.20 0.20 0.10 0.10
0.4
0.10 0.04 0.04 0.60
0.44
0.60
0.44
0.03 0.02 0.60
0.10
2
2 1
0.05 0.03 0.07 0.10 0.10 0.10 0.07 0.02 0.32 0.20 0.07 0.15 0.05 0.07
0.97 0.3
7
17 1 1
1
0.08 0.04 22
0.15 0.05
6
0.06 0.10 0.07
1.94
7.08
0.03 0.04 0.04
1
0.03 0.03 0.08 0.19
1 2 14
0.03 0.02 1.42 0.05 0.06 0.04 0.05 0.04
1
1
Egészségügyi adatvagyon hasznosítása a stabil coronaria betegek ellátásának elemzésére 2012 mo_f_cikk 5 0.06 Personalized Nutrition Counseling Expert System 2012 angol_konf_cikk 5 0.04 1 Regional differences in the utilisation of coronary angiography as initial investigation for 2012 theabstract evaluation of patients with suspected coronary artery disease The effect of waiting times on the patient pathways in the evaluation of patients with2012 suspected abstract coronary artery disease 2013 Formalizing harmony rules for nutrition counseling 2013 külf_f_cikk 3 0.13 Lokális agykérgi aktivitás mérése Laplace-típusú EEG térképezéssel: A felbontás vizsgálata 2013modellezéssel mo_f_cikk 5 0.06 Challenges and limits in personalized dietary logging and analysis 2013 angol_konf_cikk 3 0.07 Étel adatbázisok tartalmi eltéréseinek hatása a diéta naplózás pontosságára 2013 magyar_konf_cikk 4 0.03 Személyre szabott táplálkozás-tanácsadó rendszer harmónia szabályokkal 2013 magyar_konf_cikk 3 0.03 Táplálkozási adatbázisok átjárhatósága 2013 magyar_konf_cikk 2 0.05 A regionális vizsgálati frekvenciák és a viszgálatra kerülok halálozási mutatójának összefüggése 2013 magyar_konf_cikk koszorúérbetegség gyanújával 5 kivizsgálásra kerülo betegekben 0.02 Kardiológiai rehabilitációs kezelésben részesülo illetve ilyen kezelésre potenciálisan jelölt 2013betegek magyar_konf_cikk gyógyszerfogyasztásának 5 összehasonlítása 0.02 GAME: GPU accelerated multipurpose evolutionary algorithm library 2013 külf_f_cikk 4 0.10 2 Dietary Logging and Analysis for Tele-Care Using Harmony Rules 2013 angol_konf_cikk 6 0.03 SOLO: An EEG Processing Software Framework for Localising Epileptogenic Zones 2013 angol_konf_cikk 6 0.03 Életmód-változtatást támogató mobil informatikai alkalmazások 2013 mo_f_cikk 5 0.06 Epileptikus gócok EEG alapú lokalizálását támogató szoftver környezet 2013 mo_f_cikk 4 0.08 Regional differences in the utilisation of coronary angiography as initial investigation for 2013 theif_cikk evaluation of patients with suspected 5 5.5coronary artery disease 1.10 1 2014 The impact of geographical distances to coronary angiography laboratories on the patient 2014evaluation if_cikk pathways in patients4 with0.06 suspected coronary artery 0.15 disease. Results 1 from a population-based study i The Effect of the Waiting Times on the Patient Pathways for Patients with Suspected Coronary 2014 angol_konf_cikk Artery Disease 3 0.07 Combined Model for Diabetes Lifestyle Support 2014 angol_konf_cikk 3 0.07 Preclinical tests of an android based dietary logging application 2014 angol_konf_cikk 6 0.03 Lifestyle Log Based Blood Glucose Level Prediction for Outpatient Care 2014 angol_konf_cikk 3 0.07 Personalized Dietary Counseling System Using Harmony Rules in Tele-Care 2014 külf_f_cikk 4 0.10 A fast, android based dietary logging application to support the life style change of cardio-metabolic 2014 angol_konf_cikk patients 6 0.03 Képfeldolgozás folyamata az EEG neuroesztétikai vizsgálatában 2014 mo_f_cikk 4 0.08 Rehabilitációra érdemes és ténylegesen rehabilitációra kerülo betegellátási utak elemzése 2014 magyar_konf_cikk 7 0.01 Felhasználói tapasztalatok a Lavinia mobil életmód-tükör alkalmazással 2014 magyar_konf_cikk 7 0.01 Életmód-naplózás pontosságának elemzése kardiológiai rehabilitációra kerülo betegek2014 körében magyar_konf_cikk 6 0.02 Okostelefon alapú diéta-naplózó rendszer diabeteses betegek számára 2014 mo_f_cikk 4 0.08 Kamrai szívizom-repolarizáció heterogenitás modellezéses vizsgálata 2014 mo_f_cikk 3 0.10 Telemedical Heart Rate Measurements for Lifestyle Counselling 2014 mo_f_cikk 4 0.08 A szívkatéteres laboratóriumoktól mért földrajzi távolság hatása az iszkémiás szívbetegség 2014 gyanújával mo_f_cikk ellátásra került betegek 4 ellátási útjára 0.08 1 Diabétesz életmód-támogatás vércukorszint-előrejelzéssel 2014 mo_f_cikk 4 0.08 Reliability of telemedical Heart Rate meters 2014 mo_f_cikk 4 0.08 Kamrai szívizom-repolarizáció heterogenitás vizsgálat bioelektromos képalkotóval 2014 mo_f_cikk 4 0.08 Távmonitorozásra is alkalmas pitvari fibrilláció detektálási módszer 2014 mo_f_cikk 3 0.10 2015 Validation of a low cost telemedical stress monitoring system 2015 angol_konf_cikk 2 0.10 Flexibilis, eseményvezérelt keretrendszer mobil telemedicinális alkalmazásokhoz 2015 magyar_konf_cikk 5 0.02 Glikémiás hatást befolyásoló életmódbeli, étrendi tényezok vizsgálata cukorbetegek vércukor 2015 magyar_konf_cikk szintjére 6 0.02 Hosszú hatású inzulin kezelése vércukorszint-előrejelző modellben 2015 magyar_konf_cikk 3 0.03 Vércukor-előrejelző modell klinikai validációja 2015 magyar_konf_cikk 6 0.02 Étrendértékelés és -tervezés mesterséges intelligencia segítségével 2015 magyar_konf_cikk 5 0.02 Fiziológiai paraméterek változása életmód támogató informatikai rendszer használata2015 soránmagyar_konf_cikk 5 0.02 Stabil anginás betegutak klaszterelemzése 2015 magyar_konf_cikk 3 0.03 1 Automatic stress detection using simple telemedical heart rate meters 2015 angol_konf_cikk 2 0.10 Effectiveness of mobile personal dietary logging 2015 angol_konf_cikk 5 0.04 Diabetes Lifestyle Support with Improved Glycemia Prediction Algorithm 2015 angol_konf_cikk 4 0.05 Blood Glucose Level Prediction for Mobile Lifestyle Counseling 2015 angol_konf_cikk 5 0.04 2016 Táplálkozási trendek, szakmai ajánlások. Mi a jövő útja?: Életmód-támogató mobil alkalmazás 2016 mo_f_cikk szerepe a dietoterápiában 4 0.08 Diabetesesek dietoterápiájának és önmenedzselésének támogatása mobilapplikációk 2016 használatával if_cikk 4 0.29 0.15 Stress Detection Using Low Cost Heart Rate Sensors 2016 if_cikk 3 0.92 0.31 A betegek anonimitásának biztosítása a földrajzi elhelyezkedésre kiterjedő egészségügyi 2016 adatelemzések mo_f_cikk során 4 0.08 Characterizing blood glucose response to specific meals in pre-diabetes: a small scale 2016 studyangol_konf_cikk 3 0.07 Changes in the spatial distribution of dominant IHD care providers over a 10 year period 2016 in Hungary angol_konf_cikk 3 0.07 mHealth szolgáltatás felhasználói igényének felmérése 2016 magyar_konf_cikk 6 0.02 2017 Blood glucose response characterization for outpatient pre-diabetes care 2017 mo_f_cikk 3 0.10
MTA típusok if_cikk külf_f_cikk mo_f_cikk angol_konf_cikk magyar_konf_cikk könyvrészlet abstract
65
A tíz legfontosabbnak ítélt publikáció [1] I.Vassányi. Implementing fine grain processor arrays on field-programmable logic. Integrated Computer-Aided Engineering, Special Issue on Architectural Trends for Image Processing and Machine Vision, Vol. 7, Number 1, 2000 pp. 53-67. (SCI impact factor: 0.44) [2] Istvan Vassanyi, Jozsef Barcza, Tamas Tobak. Making two-level standards work in legacy communication frameworks. Proc. XIXth Int. Congress of the European Federation for Medical Informatics (MIE2005), Geneva, Switzerland, Aug-Sept. 2005, pp. 689-695. [3] B. Gaál, I. Vassányi, G. Kozmann. A Novel Artificial Intelligence Method for Weekly Dietary Menu Planning. Methods of Information in Medicine 2005; 44: 655–64. (SCI impact factor: 0.97) ISSN: 0026-1270 [4] S.Palágyi, O.Csirke, J.Futó, J.Hlavay, B.Raucsik, A.Szabó, I.Vassányi. Mining Data from Roman Sandstone Quarries. Acta Archeologica Academiae Scientiarum Hung. 57 (2006) 395-422. [5] Colombo G, Merico D, Boncoraglio G, De Paoli F, Ellul J, Frisoni G, Nagy Z, van der Lugt A, Vassányi I, Antoniotti M. An ontological modeling approach to cerebrovascular disease studies: The NEUROWEB case. Journal of Biomedical Informatics, 2010 Aug, 43(4):469-84 (SCI impact factor: 1.94) [6] I.Kósa, I.Vassányi, A.Nemes, J.Hortobágyi, Gy.Kozmann. Stress ECG utilization in the evaluation of patients with chest pain: The real practice in Hungary with 10 million inhabitants. International Journal of Cardiology, Volume 149, Issue 1, May 2011, ISSN 0167-5273, pp. 137-139. (SCI impact factor: 7.08) [7] István Kósa, Attila Nemes, Éva Belicza, Ferenc Király, István Vassányi. Regional differences in the utilisation of coronary angiography as initial investigation for the evaluation of patients with suspected coronary artery disease. International Journal of Cardiology, Volume 168, Issue 5, 12 October 2013, Pp. 5012-5015, ISSN 0167-5273, doi: 10.1016/j.ijcard.2013.07.148, (SCI impact factor: 5.5) [8] Peter Gyuk, Tamas Lorincz, Rebaz A. H. Karim, Istvan Vassanyi. Diabetes Lifestyle Support with Improved Glycemia Prediction Algorithm. In Marike Hettinga et al. (eds.) Proc. of the Seventh International Conference on eHealth, Telemedicine, and Social Medicine (eTELEMED 2015), ISBN 978-1-61208-384-1, Lisboa, Portugal, 22-27 February 2015, pp. 95-100. [9] Istvan Vassanyi, Istvan Kosa, Rebaz A.H. Karim, Marta Nemes and Brigitta Szalka. Effectiveness of mobile personal dietary logging. In Piet Kommers and Pedro Isaías (eds.) Proc. of the 13th Int. Conf. e-Society 2015, Madeira, Portugal, March 14-16, 2015, ISBN 978-989-8533-32-6, pp. 288292. [10]Mario Salai, István Vassányi, István Kósa. Stress Detection Using Low Cost Heart Rate Sensors. Journal of Healthcare Engineering, Volume 2016 (2016), 13 pages, DOI 10.1155/2016/5136705 (SCI impact factor 0.92)
66
A LEGFONTOSABBNAK ÍTÉLT 10 PUBLIKÁCIÓ KÜLÖNLENYOMATA
67
Accepted as full paper for the XIX Int. Congress of the European Federation for Medical Informatics (MIE’2005), 28 Aug-1 Sept 2005, Geneva, Switzerland.
Making two-level standards work in legacy communication frameworks Istvan Vassanyi, Jozsef Barcza, Tamas Tobak University of Veszprém, Department of Information Systems Veszprém, Hungary
Abstract The research presented deals with the problem of how information technology can support the interconnection of medical information systems in practice. We apply a two-part structural standard that consists of a fixed Information Reference (IR) Model, and an Archetype Reference (AR) Model that uses the structures of the IR model. A very important problem with the real-world application of such a two-level model is that the high level (abstract) entities, structures and their connections must be somehow translated into lower level equivalents that legacy database information systems can actually use to program their standardised interface. Our choice for the lower level medium is XML, such that the standard appears as an XSD schema that can be used, in the usual way, to validate a message document. To test the viability of the above paradigm, we developed an archetype-XSD translator tool in the form of Protégé/OWL plugins and tested it on an industrial interface for exchanging medical episodes (MedQuery), using an implementation of openEHR. We found that the most important features: Containment, Cardinality, Named references to other instances, and References to external terminologies, could all be mapped to standard XSD constructs. We also developed a validator plugin to check external references. We plan to put the system at work in a heterogeneous medical messaging system (a descendant of the Budapest based MediNet system) in the near future.
Keywords: Medical Records, Structural Mapping, Archetypes, XML
1. Introduction The interconnection of medical information systems for automated information exchange is an ultimate goal to improve cost-efficiency and quality of service in the health care domain. The difficulty is, however, that due to historic, economic and cultural reasons, the information systems in most European medical institutions, especially in Eastern Europe, were developed as a local initiative within the institution and without much effort for standardization [1]. This resulted in systems with totally different information models, even in the same area of specialization, sometimes even within the same institution. Recently, new structural standards by CEN, HL7 (v3 RIM), openEHR, etc. have appeared [2-4], and they are being applied in pilot research and industrial projects to implement new information systems e.g. [5, 6]. However, the question of today is how to interconnect heterogeneous legacy systems, based on huge relational databases. Existing system interconnection schemes can be classified into two broad categories:
1. Point-to-point programmed connection between two specific systems, with or without the application of a messaging standard (e.g. HL7). The standard, if used, does not specify the actual content, i.e. the DOCUMENT part of the message. This is the most commonly used scheme in practice, and also the most popular solution with legacy system vendors. This way a new ad-hoc interface must be specified and implemented for any two communicating parties. 2. A messaging standard is used in combination with a field level message content standard [7]. This solution means that programming a single interface for each communicating party is sufficient. However, the content standard usually has a fixed, flat–file structure and the correct interpretation of the fields depends on the programmers of the interfaces. The content standard cannot express sophisticated information structures (like that of all the data making up a cardiac episode) and does not support object references among sub-structures. To improve the shortcomings of present-day solutions, structural standards should be used for the message content [8]. All major emerging healthcare standards have extensions to build such models. Among them, openEHR, the applicable standards of the CEN, and also the recently developed Hungarian national pre-standard adopts a twolevel approach by splitting the structural model in two parts (Fig 1): • a fixed Information Reference (IR) Model, which contains a hierarchy of information structures like arrays, folders, etc., and • an Archetype Reference Model that builds upon the structures of the IR model, and that is edited and updated by a committee of medical experts. An example of an AR model is the structure and content of an inpatient episode at a hospital department. The applicability of two-level standards was proven in several studies [9, 10]. Since the IR Model is fixed, software can be developed independent of changes in Archetype Models [2]. Due to the phenomena, as well as their relations, of medical science being very complex, Archetype Models must be defined at a high level of abstraction, preferably in the form of an ontology. In such an ontology, the classes normally form the IR model, and the instances of these classes form the archetypes. The focus of the research presented in this paper was on the practical application of archetype ontologies to build communication frameworks. 2. Methods In our system model, messages from all communicating parties are sent via a data centre. The centre stores the archetype definitions, provides public access to them for all parties and for the development team of medical experts. Each message passing the centre references an archetype registered in the archetype store. The centre checks conformity with the referenced model and also checks the validity of the terminology references (for the role of terminologies, see Section 3.4).
MIS 1
M essage ( XML)
Archetype development
Reference for AT X
Info. model Archetype X
C E NT R E Host system : • data security • medical statistics • data warehouse
Translate into XSD
• AT administration • message valid ation
√
M essage ( XML)
MIS 2
Fig 1: Model of the communication system (AT: Archetype, MIS: Medical information system) Additional functions of the centre may include data security, medical data warehousing and data mining. These functions, especially data mining, contribute to the costeffectiveness of the overall health system. An overview of the system is shown in Fig. 1. An example for such a centralized medical communication system, although without support for archetypes, is the experimental MEDINET system in Budapest that connects GPs to hospitals and service providers. 3. Results The basic problem with the real-world application of an archetype is that the high level (abstract) entities and their connections must be somehow translated into a lower level equivalent that legacy systems can use to program their standardised interface for the archetype. Our choice for the lower level technology is XML. This means that the archetype appears as an XML schema that can be readily used to validate the message document. XML appears at a medium level between the archetype ontology and the relational database structures. Table 1 shows an overview of these three abstraction levels. Abstraction level
Model type
Model ownership
High
Archetype ontology
Medical standard
Medium
XML schema
IT standard
Low
Relational structures
Proprietary
Table 1: Levels of structural modelling
XML also has a wide support across various database technologies, e.g. the MS SQL Server uses XSD Mapping Schema and XML Views for the XML-relational mapping. Although XML is at a significantly lower level of abstraction than the archetype ontology, the interface programmer of the medical information system must still augment the relational data with domain-specific knowledge to produce a valid message. This is illustrated in Fig 2. Archetype Model
XML message
(ontology instances)
(Ref.to Diagnosis archetype) archetype)
Diagnosis(org .) Diagnosis(org.) contains{ diag_admin_data(obs .) diag_admin_data(obs.) hasInterface{ hasInterface{ diag_type(ret_value diag_type(ret_value)) diag_date(ret_value diag_date(ret_value)) ... } diag_content(obs .) diag_content(obs.) hasInterface{ hasInterface{ diag_code(ret_value diag_code(ret_value)) diag_termset(ret_value diag_termset(ret_value)) … } in_set{ in_set{ Requiredl(set Requiredl(set)) } … }
…
Abstract annotations
Relational schema Episode
Diag_types
ep_id …
type_id …
Diagnosis
XSD schema <xs:element ref=‘diag_contentl’ ref=‘diag_contentl’ minOccurs=‘1’ …
XML constructs
Data
diag_id ep_id diag_date diag_type diag_code …
Implicit context info. info. (programmed) programmed)
Fig 2: Example of mapping among different structural levels In the figure, the “implicit contextual information” added by the interface programmer is the correct classification of the relational data. For example, the structural standard may prescribe separate folders for administrative and medical data items that are stored mixed in the same table. Our basic goal was to examine how the crucial features of the ontology can be mapped to natural XML constructs. We identified four minimal features that the XML schema must be able to express: Containment, Cardinality, References to other instances within the message, and References to external terminologies. 3.1. Containment
Data structures in an archetype are normally nested into each other. For example, an observation called Primary diagnosis may be contained in an organizer called Anamnesis which in turn may be contained in a folder called Medical data. Containment appears in the archetype ontology as a special reference, called “Contains/Is_contained_by” between two instances within an archetype. This relation quite naturally maps into nested <element> XML structures. The nested structure makes the message easy to understand even for the human reader. 3.2. Cardinality
Cardinality means the allowable range of the number of references, contained items etc at any level. An ontology normally allows more freedom in specifying cardinality than XML which supports it only via the minOccurs/maxOccurs properties. However, our experiences show that very special cardinalities are not really necessary, nor used, in real-life archetypes. So, for this feature the basic XML construct will suffer. 3.3. References to other instances
An actual XML message references an archetype which defines its allowed structure. For example, we may prescribe that an organizer called “Diagnoses” must exist that
must contain zero or more Diagnosis observations. The archetype model may also specify that there may/must be certain references among the objects making up the archetype. For example, a diagnosis may have a possibly multi-valued reference called “Contradicting diagnoses”, pointing to other diagnoses. When using this feature, there will be more than one diagnosis instances in the same Diagnoses organizer in the actual XML message. To implement the reference as a property of the diagnosis element, we cannot use the key/keyref XML construct as this would not allow the reference to Diagnoses placed in other places of the document tree. So we had to resort to the old ID/IDref construct that requires a message-global identifier to be created for each element that can be referenced. The other, worse, solution would be placing all elements in the root of the tree, and thereby losing the visible containment structure. However, creating global identifiers from the object (archetype) name and database table identifier should not be a great pain for the interface programmers. 3.4. References to external terminologies
Term sets, code-dictionaries, hierarchical terminologies form a major part of all data contained in medical information systems. Some of these terminologies are used only locally, like the codes of the wards, some are also proprietary. The archetype must allow the message to contain all sorts of external references. XML by itself cannot guarantee the correctness and existence of an external terminology reference, which appears in the message as a simple text string. We propose that the Data Centre should validate external references instead. For this, the archetype must specify the type and URI of the referenced terminology. For proprietary or hidden terminologies, the validation may use a validator module (plugin) provided by the owner. This was the approach we took in our test implementation. 4. Discussion To test the viability of the above principles, we implemented an archetype-XSD translator tool (called Schema Maker) and a separate Validator tool as Protégé plugins in a test system. The Information Reference Model we used was the openEHR class hierarchy in the implementation of Isabel Roman Martinez [11]. For the test archetype, we chose the relatively simple MedQuery industrial messaging standard, which we manually implemented with the openEHR information reference model. This standard is used in some medical information systems in Hungary to transfer all sorts of information on medical episodes in hospitals. The standard contained examples for all four modeling issues discussed above. The tools correctly produce the XML schema for the tested cases. Why not simply use XML as the only means of defining structural standards? That way we would lose the power of the two-level approach that effectively constrains the possible archetype structures with the building blocks of the information reference model. However, it must be emphasized that our solution for the transform of the ontology into XML schema, although straightforward, is not the only one possible solution. In order for the programs to work, we needed to create and adhere to certain conventions like the configuration and naming of references in the archetype definition, or whether we prescribe the ID tags for all elements in the message that can be referenced, or only for those that are actually referenced, etc. In fact these conventions form a meta-standard on the interpretation of the ontology that would come otherwise in an ad hoc manner from the interface programmers.
5. Conclusions and future work The paper analyzed the problem of how information technology can support the standardised transfer of structural information among health information systems. Adopting the principles of the archetype-based, two-level medical standards, we identified the correct interpretation of high level Archetype Model constructs as the most crucial aspect of practical industrial implementation. In the demo system presented in the paper, we translate the archetype model ontology into a suitable middle level description (XML schema) that is already a useful and unambiguous standard for a programmer of legacy healthcare information systems, and that can easily be used for the validation of messages. We found that the most important features: Containment, Cardinality, References to other instances within the message, and References to external terminologies, could all be safely mapped to XSD constructs. We plan use the concept and the tools in a heterogeneous medical messaging system (a descendant of the Budapest based MediNet system) soon. Acknowledgements The work presented was supported by the Hungarian Scientific Research Fund, grant no. F037416 and IKTA 142/2003. References 1.
Frieder Klein. What is the present status of electronic patient’s record in Germany – a user’s personal experience. Proc. European Federation for Medical Informatics, Special Topic Conference, Munich, Germany 13-16 June 2004, pp. 87-88.
2.
The openEHR EHR Reference Model, http://www.openehr.org
3.
CEN/TC 251, http://www.centc251.org
4.
HL7 V3 Reference Information Model (Health Level Seven, Inc), http://www.hl7.org
5.
Bernd Blobel. Advanced EHR architectures – promises or reality, Proc. Eur. Federation for Medical Informatics, Special Topic Conference. 13-16 June 2004, Munich, Germany, pp. 73-78.
6.
Petr Hanzlicek, Josef Splidlen, Miroslav Nagy. Universal electronic health record MUDR, in: Duplaga M, Zielinski K, Ingram D, eds. Transformation of Healthcare with Information Technologies, IOS Press, 2004. pp. 190-201.
7.
Stergiani Spyrou, Alexander Berlerb, Panagiotis Bamidis. Information System Interoperability in a Regional Health Care System Infrastructure: a pilot study using Health Care Information Standards. Proc. XVIIIth International Congress of the European Federation for Medical Informatics (MIE’2003), Saint-Malo, France, 4-7 May 2003, pp. 364-369.
8.
Frank Oemig, Bernd Bloebel. Making Messaging Standards Work: From Definition to Interoperability at Runtime. Proc. MIE’2003, pp. 679-683.
9.
C.A.Brandt, K.Sun, P.Charpentier, P.M.Nadkarni. “Integration of Web-based and PC-based clinical research databases.” Methods Inf Med 2004; 43: 287-95.
10. Knut Bernstein, Morten Bruun-Rasmussen, Soren Vingtoft, Stig Kjar Andersen, Christian Nohr. Modelling and implementing Electronic Health Records in Denmark. Proc. MIE’2003, pp. 245-250. 11. Isabel Roman Martinez. An implementation of the openEHR Information Reference Model. at http://trajano.us.es/%7Eisabel/EHR/
Address for correspondence Istvan Vassanyi, University of Veszprem, Dept. Information Systems, Veszprem, Egyetem u. 10, H-8200, Hungary. Email:
[email protected]
A fast, android based dietary logging application to support the life style change of cardio-metabolic patients I. Kósa1,2, I. Vassányi1, M. Nemes1, K.H. Kálmánné2, B Pintér1, L. Kohut2 1 Medical Informatics R&D Center, University of Pannonia,
[email protected], Egyetem u. 10, Veszprem, Hungary, b Cardiac Rehabilitation Centre of Military Hospital, Szabadság u 4, Balatonfüred, Hungary Introduction The increased burden of chronic diseases in the health care of modern societies is well known. It is clear that old methods to cover the health supply demand of the society are insufficient, so services utilising modern info-communication technologies are taking over a part of classic workflows of the chronic disease management [1, 2]. The majority of these chronic diseases root, however, in a life style deviation, so systems supporting only the remote measurement of physiological are insufficient, because they miss the very important intervention possibilities of life style modification. The correction of life style deviations is one of the most human effort intensive procedures in the health care. However, it can also be supported by modern info-communication technologies. The use of a dietary log application to support later dietary counselling is a typical function where apps of mobile devices can be very effective. The basis of most computerized services available for dietary logging is a Food Composition Database (FCDB) containing the nutrient content for a wide range of common foods. Nutrients are the basic carbohydrates, proteins, minerals, vitamins etc. and foods are either ingredients of dishes (like white flour or olive oil) or they are consumed alone like apple or red wine. Some FCDB’s, like the USDA SR FCDB are free for download [3]. In order to compute the nutrient content of the user’s meals, the dietary database must also contain a highly culture-specific set of recipes as well. There are several web-based or mobile applications that provide an interface for the logging of the user’s meal and various services related to the analysis of the log like daily/weekly overview charts, support for losing or gaining weight etc. The Calorie Counter android application boasts with the biggest recipe database of more than half a million [4], a large part of which was contributed by the users. Similar applications, with a smaller database, are the My Diet Diary and Tracker 2 Go [5, 6]. Our research group at the
University of Pannonia, Hungary, earlier developed an android based dietary logging application called “Lavinia”. In a previous study we have already tested the ability of this system to cover the diet of an inpatient cardiac rehabilitation facility working with five parallel arrays of menu for a three week rehabilitation treatment [7]. Objective: The purpose of the current study is to test the time demand that the users need for dietary logging with the “Lavinia lifestyle mirror” application. Short depiction of the Lavinia logging application The Lavinia dietary logging application was developed at the Medical Informatics R&D Centre at the University of Pannonia, Veszprem, Hungary, to support dietary logging on mobile interfaces, dietary log analysis, and also personalized menu generation using a dietary database specialized for the Hungarian culture. Its data base currently stores 9500 food items along with their nutrient contents and 1373 dishes composed from the foods, but on the android user interface we show only the most important 299 dishes and 360 foods, organized in 195 sets, to simplify the search and thus preserve user motivation. The system supports a hierarchical set-based search (see Fig. 1) as well as the usual keyword based search. We also allow to log generalized dish sets (like ‘pasta’) when the user does not find an exact match. If such a set is logged, we use the averaged nutrient contents of the set members in the analysis. Methods This study was performed using the dietary data of a medical institution, the Cardiac Rehabilitation Center of the Military Hospital, Balatonfüred, Hungary. In the current, preclinical phase only staff members were included to perform a final test before the planned clinical test of the system, in which typical patients treated in the institution will take part. The test period was 22 days long, for 5 different, manually compiled regular hospital menus, so in total we processed the meals for 110 days, with 3 meals per day, with Lavinia. The full list contained 1179 logged items with 156 different dishes and 38 foods (total 194). Some items occurred quite frequently, but there were also dishes specifically designed for a certain diet.
Figure 1. Screen shots of a “set-based” search procedure in the Lavinia mobile application
The time expenditure of the mobile phone based logging procedure was measured using these 1179 items. Five persons, familiar with mobile phone usage, but new to our set based dietary logging user interface received a short (3 to 5 minutes) introduction to the operations in the Lavinia system. Then they entered the list in the Lavinia system. The activity of the test subjects was logged and time stamped by the application from the start of any new item entry to the completion of the process. The switch from the set based search to the free text based search was allowed at any point for the subjects. We measured the average time spent for an item at the beginning and end of the process. The application start up time was not considered, assuming continuous stand by position for the investigated application in the background on the android system. Results The five persons recorded the institutional menus in Lavinia for the investigated 22 days with an average of 12.0 items per day. The average net time consumption for entering a single item decreased considerably from 25.60 sec on the first day to 12.45 sec on the last, with a typical logarithmical learning curve (time=-4,06*ln(days)+26,95, R2=0.60). This decrease was dominated by the acceleration of the set based recording from
24.89 sec to 12.00 sec (time=-3,52*ln(days)+22,71, R2=0.75), while neither the frequency nor the time consumption of keyword based search was changed during this short period (13,4%, extra 22.15 sec.). The overall result is that the average total daily time consumption of dietary logging decreased from 6.80 min to 2.61 min (time=-0,86*ln(days)+5,49, R2=0.65). Conclusion: The set based dietary logging application is a viable system to generate a nutrition mirror for the users. The daily total time consumption of dietary logging is highly acceptable. Users possibly need longer practice to reduce the extra efforts connected with keyword based search Acknowledgment The work presented was supported by the European Union and cofunded by the European Social Fund, project title: “Telemedicine-focused research activities in the field of Mathematics, Informatics and Medical Sciences”, project number: TÁMOP-4.2.2.A-11/1/KONV-2012-0073. References [1] [2]
[3] [4] [5] [6] [7]
Cherry JC, Moffatt TP, Rodriguez C, Dryden K. Diabetes disease management program for an indigent population empowered by telemedicine technology. Diabetes Technol Ther 2002;4:783–91. Weintraub A, Gregory D, Patel AR, Levine D, Venesy D, Perry K, et al. A Multicenter Randomized Controlled Evaluation of Automated Home Monitoring and Telephonic Disease Management in Patients Recently Hospitalized for Congestive Heart Failure: The SPAN-CHF II Trial. J Card Fail 2010;16:285–92. USDA National Nutrient Database. http://www.ars.usda.gov/main/main.htm. Calorie Counter PRO MyNetDiary (verified Sept. 2013) https://play.google.com/store/apps/details?id=com.fourtechnologies.mynetdiary.ad My Diet Diary. (verified Sept. 2013) https://play.google.com/store/apps/details?id=org.medhelp.mydiet Tracker 2 Go. (verified Sept. 2013) https://play.google.com/store/apps/details?id=com.byoni.tracker2go I. Vassanyi, I. Kosa, B.Pinter. Challenges and limits in personalized dietary logging and analysis. Proc. 17th Int. Conf. on Biomedical Engineering, 28-29 November 2013, Kaunas, Lithuania, pp 22-25..
655 © 2005
Schattauer GmbH
A Novel Artificial Intelligence Method for Weekly Dietary Menu Planning B. Gaál, I. Vassányi, G. Kozmann Department of Information Systems, University of Veszprém, Hungary
Summary
Objectives: Menu planning is an important part of personalized lifestyle counseling. The paper describes the results of an automated menu generator (MenuGene) of the web-based lifestyle counseling system Cordelia that provides personalized advice to prevent cardiovascular diseases. Methods: The menu generator uses genetic algorithms to prepare weekly menus for web users. The objectives are derived from personal medical data collected via forms in Cordelia, combined with general nutritional guidelines. The weekly menu is modeled as a multilevel structure. Results: Results show that the genetic algorithm-based method succeeds in planning dietary menus that satisfy strict numerical constraints on every nutritional level (meal, daily basis, weekly basis). The rule-based assessment proved capable of manipulating the mean occurrence of the nutritional components thus providing a method for adjusting the variety and harmony of the menu plans. Conclusions: By splitting the problem into well determined sub-problems, weekly menu plans that satisfy nutritional constraints and have well assorted components can be generated with the same method that is for daily and meal plan generation.
Keywords
Genetic algorithms, multi-objective optimization, nutrition counseling Methods Inf Med 2005; 44: 655–64
Received: July 29, 2004; accepted: May 16, 2005
1. Introduction The Internet is a common medium for lifestyle counseling systems. Most systems provide only general advice in a particular field; others employ forms to categorize the user in order to give more specific information. They also often contain interactive tools for menu planning [1]. The aim of the Cordelia project [2] is to promote the prevention of cardiovascular diseases (CD), identified as the leading cause of death in Hungary, by providing personalized advice on various aspects of lifestyle, an important part of which is nutrition. MenuGene, the automated menu planner integrated with Cordelia uses the computational potential of today’s computers, which offers algorithmic solutions to hard problems. The quality of these computer-made solutions may be lower than those of qualified human professionals, but they can be computed on demand and in unlimited quantities. Nutrition counseling is one of these kinds of problems. Human professionals possibly surpass computer algorithms in quality, although research comparing performance has been ongoing since the 1960’s. In 1964 Balintfy developed a linear programming method for optimizing menus [3] and Eckstein used random search to satisfy nutrient constraints [4]. Later, artificial intelligence methods were developed mostly using Case-Based Reasoning (CBR) or Rule-Based Reasoning (RBR) or combining the two with other techniques [5]. A hybrid CBR-RBR system CAMPER [6] integrates the advantages of the two independent implementations: the case-based menu planner, CAMP [7] and PRISM [8]. A more recent CBR approach is MIKAS, menu construction using incremental knowledge ac-
quisition system [9]. MIKAS allows the incremental development of its knowledge base. Whenever the results are unsatisfactory, an expert will manually modify the system-produced diet [10]. A web-based system that models the workflow of dietitians has recently been built in Malaysia for dietary menu generation and management [11]. The core idea of our algorithm is the hierarchical organization and parallel solution of the problem. Through the decomposition of the weekly menu planning problem, nutrient constraints can simultaneously be satisfied on the level of meals, daily plans and weekly plans. This feature, which is a novelty, makes the implementation of our method instantly applicable in practice.
2. Objectives There is no generally accepted method for producing a good menu. Additionally, a menu plan, whether it is weekly, daily or single meal, can only be evaluated when it is fully constructed. So the basic objective of our work is to design a menu planner that also includes some method to evaluate menu plans.
2.1 Evaluation of Menu Plans The evaluation of a meal plan has at least two aspects. Firstly, we must consider the quantity of nutrients. There are well defined constraints for the intake of nutrient components such as carbohydrates, fat or protein which can be computed for everybody, given his/her age, gender, body mass, type of work, age and diseases. Optimal and extreme values can be specified for each nutrient component. So as for quantity, the task Methods Inf Med 5/2005
656 Gaál et al.
of planning a meal can be formulated as a constraint satisfaction and optimization problem. Secondly, the harmony of the meal’s components should be considered. Plans satisfying nutritional constraints should also be appetizing. The dishes of a meal should go together. By common sense some dishes or nutrients do not appeal in the way others do. This common sense of taste and cuisine can be described by simple rules recording the components that should fit together. There could be conflicting numerical constraints or harmony rules. A study found that menus made by professionals may fail to satisfy all of the nutrient constraints [12].
2.2 Calculation of Personalized Objectives The information collected via web forms in Cordelia explores controllable and uncontrollable risk factors for CD. Controllable risk factors include smoking, high blood pressure, diabetes, high cholesterol level, obesitas, lack of physical activity, stress and oral contraceptives. Uncontrollable factors considered are age, gender and family CD history. Based on the answers, the user is classified, the classification being a combination of factors like weight, high cholesterol, etc. MenuGene uses the user’s classification and all other useful observations (like the gender) and personal preferences (set by the user) to plan daily and weekly menus. This information is used to design the actual runtime parameters (objectives) of the menu to be generated when MenuGene is run. The nutritional allowances are looked-up from a
table similar to Dietary Reference Intakes (DRI) [13, 14]. The fact base of MenuGene was loaded with the data of a commercial nutritional database, developed especially for Hungarian lifestyle and cuisine, that at present contains the recipes of 569 dishes with 1054 ingredients. The database stores the nutritional composition of the ingredients. The recipes specify the quantity of each ingredient in the meal, so the nutrients of a meal can be calculated by summation. At present, the nutrients contained in the database for each ingredient are energy, protein, fat, carbohydrates, fiber, salt, water, and potassium. Additionally, the database contains the categorization of the ingredient as either of the following: cereal, vegetable, fruit, dairy, meat or egg, fat and candy. This classification is used by MenuGene to check whether the overall composition (with respect to the ratio of the categories) conforms to the recommendations of the “food pyramid”.
3. Methods MenuGene uses genetic algorithms for the generation of dietary plans. A genetic algorithm (GA) is an algorithm used for the solution of difficult problems by the application of the principles of evolutionary biology and computer science. Genetic algorithms use techniques such as inheritance, mutation, natural selection and recombination derived from biology. In GAs a population of abstract representations of candidate solutions (also called chromosomes, genomes or individuals) evolves toward better solutions. The evolution starts with a population containing random indi-
Table 1 The crossover operator is shown in the table in function of the nutritional level. Legend: M – Monday, Tu – Tuesday, W – Wednesday, Th – Thursday, F – Friday, Sa – Saturday, Su – Sunday, BF – breakfast, MS – morning snack, L – lunch, AS – afternoon snack, D – dinner, S – soup, G – garnish, T – topping, Dr – drink, De – dessert, cp – crossover point
Methods Inf Med 5/2005
viduals and happens in generations, in which stochastically selected individuals are modified (via recombination or mutation) to form the population of the next iteration. The attributes (also called alleles) of the chromosomes contain the information where each attribute represents a property. Genetic algorithms are used widely in the medical field [15-18]. GAs showed their strength in satisfying optimization problems; therefore we examined their efficiency in the generation of meal plans meeting quantitative nutrition constraints. Test software was developed to analyze the adequacy of GAs. Experiments are highlighted in the results section.
3.1 Evolutionary Operators In order to start the search process, we first need an initial population. This may be created randomly or may be loaded from a database containing solutions of similar cases (Case Based Reasoning). The population in our tests contained 40 to 200 individuals, which are meals if we plan a single meal, daily menus if we plan a daily menu etc. In the case of a meal plan, the population contains meals, the attributes of which are dishes. Then, in each iteration step, we perform a sequence of the evolutionary operators (crossover, mutation, selection) on the individuals. We stop the evolution process after a maximum of 1000 cycles or when no significant improvement could be achieved. The best individual of the final population is selected as the solution. The evolutionary operators are presented through the following example. A regular Hungarian lunch consists of five parts: 1) a soup, a main dish of 2) a garnish (e.g. mashed potatoes) and 3) a topping (e.g. a slice of meat), 4) a drink, and 5) a dessert. So, a solution for a regular Hungarian lunch contains five attributes. The crossover (also called recombination) operator involves two solutions and it means that starting at a random point, their attributes are swapped. For example, if the starting point is the last attribute, then crossover means the exchange of the desserts. Mutation replaces a randomly selected dish with another one of the right sort (e.g. a soup with another soup).
657 MenuGene – Personalized Nutrition Counseling
The single point crossover (recombination) operator is shown in Table 1. Recombination is done by randomly choosing a crossover point (cp) and creating two offsprings by exchanging the attributes of the solutions from that point on. On the weekly level, the attributes of the solutions represent daily menu plans. In the example in Table 1 we apply crossover to weekly level solutions, with a randomly chosen crossover point (cp = 3). The first offspring will contain the daily menu plans for Monday, Tuesday and Wednesday from the first parent and Thursday, Friday, Saturday and Sunday from the second parent. The genetic operators work on the abstract solution and attribute classes and do not operate on problem-specific data, thus the same method is used on every level.
3.2 Fitness Function Whenever a new individual (offspring) is created by mutation or recombination, the fitness function assesses it according to its goodness. As for numerical constraints, the fitness function has to filter out solutions having an inadequate amount of nutrients. The fitness of a solution is defined by the sums of functions composed of four quadratic curves that take their maximum (0) at the specified optimum parameters, and break down abruptly over the upper and lower limit parameters (see Fig. 1.). For example, if a set of constraints (upper and lower limit, optimal value) is separately defined for carbohydrates and proteins, then the fitness is a sum of the two values taken from the two penalty functions for the carbohydrate and protein curves at the respective amounts. The actual fitness value bears no concrete physical meaning; it is used only for comparison and selection. The individuals with the highest (i.e. closest to zero) fitness values are considered the best solutions for the search problem. The penalty function is thus designed that it should not differentiate small deviations from the optimum but be strict on values that are not in the interval defined by the constraints. The function is non-symmetric to the optimum, because the effects of the deviations from the optimal value can
be different in the negative and positive case. These penalties have been derived from the manual assessment methods of our nutritional expert. This sort of penaltybased fitness function is also often applied in other multi-objective optimization techniques [19]. After the goodness is computed as a function of the numerical constraints, the fitness function examines whether the attributes of the solution are well assorted. Rules are used for classification according to the harmony of the components. Each rule has two parts: conditions and fitness modification value. The general form of the rules is ri = 〈condition1, ..., conditionn, fitness modification value〉. The fitness value (which is less or equal to zero) should be divided by the modification value so that if less than one, the fitness will decrease. Our nutrition expert organizes every food in our database into sets. The structure of these sets are very similar to those of PRISM [8]. 〈meat〉 → 〈white meat〉|〈red meat〉 〈white meat〉 → 〈chicken〉|〈fish〉, 〈red meat〉 → 〈beef〉|〈porc〉, 〈lunch〉 → 〈lunch_with_white_meat〉|〈lunch_ with_red_ meat〉|〈vegetarian_lunch〉 The conditions part of the rules on the level of meals contains one or more dish sets (e.g. dry top) or specific dishes (e.g. tomato soup). Some example rules might look like these:
r1 = 〈dry top, dry garnish, 0.75〉, r2 = 〈tomato soup, tomato drink, 0.6〉, r3 = 〈dry top, dry garnish, pickles, 0.8〉, r4 = 〈candy, 0.7〉 Rule r2 means that for each meal that contains tomato soup and tomato drink the system will replace the fitness with 60% of the original value. Rule r3 penalizes the simultaneous occurrence of three dishes while rule r4 penalizes solutions with any kind of candy. Rules may be applied and configured for any level (e.g. daily level, meal level) of the algorithm. So, if the above rule r2 is applied at the daily level, it will reduce the fitness to 60% of those daily menus that contain tomato soup and tomato drink anywhere in their meals. Only the most appropriate rules are applied. For example, if we have a meal plan that contains tomato soup and tomato drink then from the rules r2 = 〈tomato soup, tomato drink, 0.6〉 and r5 = 〈soups, tomato drink, 0.72〉 only the former is applied because it is more specific. Only the strictest rule is applied from two or more rules with the same condition parts.
3.3 Divide and Conquer There are several common features in the different nutritional levels (meal plan, daily plan, weekly plan). For example, both a meal and a daily plan can be considered a solution of a GA, where the attributes of daily plans are meals and the attributes of meals are dishes (see Fig. 2.). The problem
Fig. 1 The Fitness function with optimum = 300, lower limit = 100, upper limit = 700
Methods Inf Med 5/2005
658 Gaál et al.
spectively of national cuisine, eating habits and nutrition database. For example, in Hungarian hospitals, “hot dinners” containing at least a hot soup are served two times a week.A solution type for this kind of weekly plan can be easily recorded in our database. It would contain five “regular daily plan with cold dinner” attributes for every day except Thursday and Sunday where there would be attributes for “regular daily plan with hot dinner”.Another example would be a school cafeteria, where only breakfasts, morning snacks and lunches are served. Five attributes for “daily plan for school refectories” would make up the solution for a weekly plan (only for weekdays).
3.4. Simultaneous Generation of Different Level Menu Plans
Fig. 2
The multi-level structure of the algorithm. Concrete examples are in square brackets.
of generating weekly menus can be divided into sub-problems, which in turn can be solved the same way using GAs. Recently a similar approach, a multi-level GA, was presented and tested on a multi-objective optimization problem [20]. For solving the problem, we created a C++ framework called GSLib, which uses the functions provided by the GALib [21] genetic algorithm library for running a standard, parameterized evolution process on the current population. GSLib is abstracted from the menu generation problem and uses abstract Methods Inf Med 5/2005
classes such as “solution” and “attribute” to represent the information related to the optimization and constraint satisfaction problems. The role of GSLib is the algorithmic setup, and the multi-level divide/conquer style scheduling and operation of the abstract evolution processes at various levels in the test system Menugene. For a description of possible scheduling strategies see Section 3.4. Because of the abstract framework, every kind of meal can be represented and every kind of plan can be generated irre-
The generation of weekly menu plans could be done in sequential form. Meal level GAs could create meals from dishes and these meals could then be used as a fact base to generate daily plans. The same applies for weekly menu plan generation. However, our method can run the different level GAs in a concerted manner. Table 2 gives an overview of the currently implemented mutation-based GA scheduling strategy and other alternatives as well. Columns 4 and 5 of Table 2 show the sequence of how GSLib fires the evolution processes on the various levels for the topdown and bottom-up strategies, respectively. For example, the top-down strategy starts with an initial population (loaded randomly or from the case-base) and goes on by evolving the weekly level for a given number of iteration steps (multi-level iteration step: 1). Then, the evolution proceeds on the level of daily plans (2) by evolving the attributes of the weekly menu plans. After the second level, the process continues on the level of meals (3), and after that, the evolution restarts from the weekly level (4). The multi-level scheduling for credit propagation and mutation-based strategy are also described in Table 2. In any strategy, there is no evolution on the level of dishes, nourishments, and nourishment components. The estimated number of instances of the solution and attribute classes is shown in
659 MenuGene – Personalized Nutrition Counseling
Table 2
The multi-level GA scheduling strategies (top-down, bottom-up, credit propagation, mutation-based) in function of algorithmic levels
Level Weekly
Daily Meal Dish Nourishment Nourishment component
Object Population
Number of instances
Top-down
Bottom-up
1
Sol
40
Attr
280
Sol
14000
Attr
70000
Sol
84e+05
Attr
336e+05
Sol
756e+07
Attr
6804e+07
Sol
68e+09
Attr
544e+09
Sol
544e+09
1.
4.
7.
3.
6.
Mutation-based
Credit propagation
9.
Start with a defined amount of credit and decide on each level what to use it for 1. for evolving the current level
2.
5.
8.
2.
5.
8.
2. use part of the credit and share the other part among the lower level objects
3.
6.
9.
1.
4.
7.
3. share all of the credit among the lower level objects
column 3. Note that by exploiting a copyon-write technique, most instances of the classes are virtual and stored in the same memory location. In contrast to the sequential method, we keep all of the adjustable parameters of the various levels in the memory to provide a larger search space in which generally better solutions can be found. So, if a GA evolving daily menu plans can’t satisfy its constraints and rules, its fact base (which consists of meals) can be improved by further evolving the populations on the level of meals.
3.5 Case-based Initialization of Initial Populations Solutions found for a given problem (for example: low cholesterol breakfast plans) can be reused by loading them to the appropriate initial populations. We therefore store some of the best solutions found for each sub-problem. Whenever Menugene begins to create a new plan which should satisfy some constraints, it searches its database for solutions that were generated with similar constraints and loads them as an initial/ startup population.
Fire evolution on lower levels when mutation occurs
normal mutation
No evolutionary process on these levels
4. Results 4.1 Optimal Crossover and Mutation Rates We performed tests to explore the best algorithmic setup. First we analyzed the effect of the crossover and mutation probabilities on fitness. We used the same randomly generated initial populations for the tests, and averaged the results of ten runs in each configuration. The results showed that while the probability of the crossover does not influence the fitness too much, a mutation rate well above 10% is desirable, particularly for
smaller populations. This result is surprising at first, as the literature of GA generally does not recommend mutation rates above 0.1 … 0.5%. However, due to the relatively large number of possible alleles, we need high mutation rates to ensure that all candidate alleles are actually considered in the evolution process.
4.2 Fast Convergence to Best Available Solution Runtime performance is an important factor for MenuGene as it must run as an on-line service in Cordelia. Runtime is determined by the number of operations to be per-
Fig. 3 Runtime of MenuGene in various algorithmic setups. Each dot represents the result of a test run.
Methods Inf Med 5/2005
660 Gaál et al.
Fig. 4 Distance from the optimum for the worst (upper curve), average (middle curve) and best (lower curve) solution of ten runs, in the function of the number of iterations.
formed in each generation. Not surprisingly, we observed a strong linear connection between the probabilities of mutation and crossover, and runtime. However, our main concern is the quality of the solution. So we examined the connection between the runtime and the quality of the solution in a wide range of algorithmic setups (adjusted parameters were number of iteration, population size, probability of mutation and crossover). As Figure 3 shows, although the quality of the solutions improves with longer runtimes (whether it is the effect of any of the adjusted parameters), the pace of improvement is very slow after a certain time, and, on the other hand, a solution of quite good quality is produced within this time. This means that it is enough to run the algorithm until the envelope curve starts to saturate. The convergence of the algorithm was measured with test runs generating meal plans. Constraints were energy (min = 4190 kJ, opt = 4200 kJ, max = 4210 kJ) and protein (29 g, 29.27 g, 29.5 g), population size was 200, and probability of crossover and mutation was 0.9 and 0.2. Results show that with two constraints, a satisfactory solution was found in 6.2 iterations on average (14 iterations in the worst case). As Figure 4 shows, there is hardly any improvement in the quality of the solution after 250 iterations, so a nearly optimal plan can be found in this time. Methods Inf Med 5/2005
4.3 Reaction to the Gradual Diminution of Constraints We tested the reaction of the algorithm to the gradual diminution of constraints. Minimal and maximal values were two times the size of the suggested at the start and were gradually decreased to be virtually equal from the aspect of human nutrition. More than 150,000 tests were run. The tests showed that our method is capable of generating nutritional plans, even where the minimal and maximal allowed values of one or two constraints are virtually equal and the algorithm finds a nearly optimal solution when there are three or four constraints of this kind. According to our nutritionist, there is no need for constraints with virtually equal minimal and maximal values, and in most pathological cases the strict regulation of four parameters is sufficient. Our method has been proven capable of generating menus with meal plans that satisfy all constraints for non-pathological nutrition.
4.4 Results of the Multi-level Method Generated plans satisfy numerical constraints on the level of meals, daily plans and weekly plans. The multi-level generation was tested with random and real world data.
The tests showed that for a mainstream desktop personal computer it takes between ten and fifteen minutes to generate a weekly menu plan with a randomly initialized population. The weekly menus satisfied numerical constraints on the level of meals, daily plans and weekly plans. Our tests showed that the rulebased classification method successfully omits components that don’t go well together. The case-based initialization of the startup population increases the speed of the generation process. Whenever a solution is needed for a plan with constraints that has been made previously it would be enough to use the solution that can be found in the case-base for these constraints. However, with some iteration, the algorithm may find better solutions than are in its initial population at startup. If there was no improvement in the best solution stored in the database for a particular plan, it can be assumed that one of the best solutions was found for that menu plan.
4.5 Variety Variety is a key factor when considering dietary plans. GAs use random choice for guiding the evolution process for near-optimum search, so if the search space is large enough, the solution found should be close to the optimum, but need not be similar in several consecutive runs. However, GAs are known for finding near-optimal solutions, so if there are strict numerical constraints, then it can easily happen that only a small subset of solutions satisfies them (which are close enough to the optimum), and the probability that the solutions don’t have similar attributes is marginal. So, a method for adjusting the expected occurrence of the alleles of the GA is needed for providing sufficient variety in the menu plans. We measured the variety of menu plans with constraints for regular dietary plans for women aged between 19 and 31 with mental occupation. The variety of the allele that represents one of the 150 possible soups for a regular lunch is shown in Figure 5. The figure shows the occurrence (ordered by frequency) of each of the 120 soups that were present more than 15 times (0.1%) in the best solutions in 15,000 test runs. The
661 MenuGene – Personalized Nutrition Counseling
most frequent soup in the best solutions of 15,000 runs was present 482 times (~3.2%), the second 426 times (~2.8%) and the 50th 102 times (~0.7%). Figure 5. also shows (lower part) the goodness of the best solution to which the corresponding alleles belong. The goodness of an allele is computed by summing its best fitness (i.e. the best fitness value of all solutions the allele was part of) with the weighted best fitness values of its eight neighbors. The goodness of the ith allele is defined as: g[i] = f [i] +
[(1 – 0.2 j) · (f [i – j] + f [i – j])],
when f [i] is the fitness of the ith allele. The trend curve (lower part) shows that solutions containing frequent alleles generally have better goodness. The results show that the algorithm uses alleles appearing in good solutions more often and the frequency of usage is approximately inversely proportional to the fitness of the best solution generated by using the particular allele. However, it may happen that a run of the algorithm with properly configured nutritional constraints and rules results in a dietary plan that contains several occurrences of same dishes or dishes made from the same ingredients. Therefore, we allow more general rules, like rs = 〈?, ?, 0.5〉 to be recorded in our rule-base, which also get pre-processed during the initialization of the algorithm. Rule rs will penalize every solution that has the same value (solution) represented by its attributes more than one time. So, if rs is imposed on a daily menu plan which contains orange drink for breakfast and lunch as well, then the fitness of this daily plan will be reduced by 50%. We measured the effect of the rules on the variety and mean occurrence of the alleles (drinks) considering the solutions for the meal plan (lunch). The results of the statistical analysis are shown in Table 3. Two rules (rA, rB) where imposed on two alleles, respectively. The strictness of the rules was decreased from 100% to 75%, 50% and finally to 25%, giving a total of 16 configurations. Rule rA penalized the solutions that contained drink “A” while rB penalized solutions with drink “B”.The relative occurrences of “A” are shown in the function of the strictness of the rules in Figure 6.
Fig. 5 The occurrences (ordered by relative frequency, upper figure) of 120 of the 150 possible alleles (soups) in the best solutions (lunches) of 15,000 runs. The lower figure shows the goodness of the best solution which the respective soup was part of. Table 3 Statistical analysis of the distribution of the potential alleles (drinks) in the best solutions (lunches) and the mean occurrence of the alleles (A, B) on which the rules were imposed.
Methods Inf Med 5/2005
662 Gaál et al.
Fig. 6 The relative occurrence of a particular solution (A) in function of the strictness of two rules (penalizing solutions A and B).
We employed a two-sample Kolmogorov-Smirnov goodness-of-fit hypothesis test with the significance level of 5% to the two random samples created by recording the alleles representing drinks in neighboring configurations, running 1000 times each (using ten random populations, running each one 100 times). The table lists those P values (denoted with KS) for which the Kolmogorov-Smirnov test has shown significant difference in the distribution of the two independent samples.
We observed significant differences for all of the test pairs with respect to increasing the strictness of rA (rows in Table 3). However, the same was not true for all of the pairs with respect to increasing the strictness of rB (columns in Table 3.), so P values are not listed for such pairs. The explanation of this phenomenon is hidden in the differences between the occurrences of alleles A and B without penalties, which are 436 (43.6%) for “A” and 68 (6.8%) for “B”, out of the 1000 possible. Since the number of
instances of “A” is comparable to the possible instances, rule rA not only changes the mean occurrence of “A”, but significantly changes the distribution of the alleles. In case of penalizing the meals with drink “A” by 75%, the occurrence count of “A” decreases by 133 (~30%) from 436 (100%) to 303 (~70%). As Figure 7 shows, the 103 occurrences are shared somewhat proportional among the other alleles (“A” is 15th, “B” is the 12th allele in Fig. 7). We performed a single sample Lilliefors hypothesis test of composite normality on the samples with an element size of 10 on 100 runs of the algorithm with 10 different starting populations and counting the occurrences of “A” and “B”. The distribution of the occurrences of “A” in function of the starting populations proved normal, except for one case. Again, due to the few occurrences of “B” in the test runs, we could not determine its distribution. We paired the samples of neighboring configurations and if both had normal distribution with a significance level of 5%, we employed paired t-tests to check, if there is a significant difference in the mean occurrences of alleles “A” and “B”. The results of the paired t-tests are shown in Table 3, denoted with T. In case of significant differences, the corresponding P values are also listed. If one of the samples did not have a normal distribution we marked the case with an asterisk (*). Since the sample distribution was not known for more than half of the samples, we employed the Wilcoxon signed rank test of equality of medians with the significance level of 5% on each sample pair to measure whether there is significant difference between the mean occurrences. Results are shown in Table 3 with the corresponding P values, and are denoted with Sr. There were only three situations where there was no significant difference between the mean occurrences. These cases are marked with a hyphen (-).
5. Discussion Fig. 7 Occurrence counts of the 15 possible alleles (drinks) in a solution (for lunch) in 1000 runs in function of the strictness of the rule imposed on the 15th allele
Methods Inf Med 5/2005
The single level tests showed that GAs are capable of generating menu plans that satisfy strict nutritional constraints. The tests
663 MenuGene – Personalized Nutrition Counseling
performed with the multi-level implementation show that constraint satisfaction remains the same on every level. From the aspect of nutritional constraints, our method outperforms present-day nutrition planning systems. Recently developed nutrition planning systems such as CAMPER [6] maintain the constraints only on a daily basis, in contrast, Menugene satisfies constraints on the meal-by-meal, daily and weekly levels. The high optimal mutation rate can be surprising at first. However, the optimal mutation probability should be high because a population with 100 or 200 individuals can’t contain much genetic flavor in the case of the menu generation problem. Mutation is the only operator which can introduce new genetic information in the population. Until solutions with proper genetic parameters appear in the population, the relatively high mutation probability is needed. It is not necessary to generate the parameters of a nutritional plan simultaneously. If one or more parameters have been previously set, the unassigned ones can be generated in a way that the whole plan satisfies the relevant constraints. A practical application of this feature is when one eats at his/her workplace and can’t choose his/her lunch for the weekdays. In this case, the lunches of the weekly plan can be defined by the end user at the beginning of the week and MenuGene can develop the whole weekly menu plan without changing these lunches. Parameters of the plan can be changed at any time; the algorithm is capable of adjusting the weekly plan to compensate for deviations. From a quantitative point of view, it is more important to keep the nutritional constraints on a weekly basis than to keep them on a daily basis or in a meal. For this reason, MenuGene allows relatively more deviation from the optimal values on the lower levels (day/meal), and tries to keep the strict constraints on a weekly basis. The advantage of our approach is that it uses the same algorithm on every level, thus the hierarchical structure is easily expandable. The method is capable of controlling the nutrition on longer periods. Monthly optimizations could be performed without the need to plan the whole monthly plan in one
run.After the first week, the plan for the second week can be made with the previous weekly plan in mind. Harmony is more important on lower levels. For example, a meal or a daily plan with two dishes or meals with tomato is not well assorted. These plans can successfully be omitted with simple rules. We developed the data structure of MenuGene in a way that it can distinguish rules loaded by experts and users. In this way, users can define their personalized rules. These rules are only used for the user that defined them and have lower priority than the rules given by experts. As MenuGene uses the rules as parameters, the rule-base of the system can be developed while the system is being used. Incremental development of the rule-base is similar to that implemented in MIKAS [8]. The increasing number of rules doesn’t have an impact on the generation time of a plan because the rule-base of the system is preprocessed.
6. Conclusions The paper described the results of the automatic, parameterized menu planner MenuGene that uses a multi-level multi-objective genetic algorithm for a near-optimum search. A genetic algorithm-based method for weekly dietary plan design was presented with a fitness function that classifies menus according to the amount of nutrients and harmony. An abstract scheme was proposed for the consistent handling of the different level problems, for the implementation of crossover and mutation and for the coding of chromosomes as well as a fitness function. Algorithmic tests revealed that a relatively high mutation rate is desirable and that after a certain time, the quality of the solution does not improve much. Our tests showed that GAs generally produce high variety, at least in non-overparameterized configurations, but in any case, rules can be used to employ the desired level of variety and harmony. Future work on MenuGene includes the development of parallel computation methods for improving the runtime of the coevolution process and the continuous im-
provement of MenuGene’s case-base and rule-base with a web-based application that was developed for human experts. The system can be tested at http://menugene.irt. vein.hu. Acknowledgments The work presented was supported by National Research and Development Program NKFP 2/052/2001 and the Hungarian Ministry of Health.
References
1. The interactive menu planner of the National Heart, Lung, and Blood Institute at http://hin. nhlbi.nih.gov/menuplanner/ [Verified April 11, 2005] 2. The Cordelia Dietary and Lifestyle counseling project at http://cordelia.vein.hu/ [Verified April 11, 2005] 3. Balintfy JL.Menu Planning by Computer. Communications of the ACM, vol. 7, no. 4, pp 255-9. April, 1964 4. Eckstein EF. Menu planning by computer: the random approach. J Am Diet Assoc 1967; 51 (6): 529-33. 5. Hinrichs RR. Problem Solving in Open Worlds: A Case Study in Design. Northvale, NJ: Erlbaum; 1992. 6. Marling CR, Petot GJ, Sterling LS. Integrating Case-Based and Rule-Based Reasoning to Meet Multiple Design Constraints. Computational Intelligenc 1999; 15: 308-12. 7. GJ Petot , CR Marling, Sterling L. An artificial intelligence system for computer-assisted menu planning. Journal of the American Dietetic Association1998; 98: 1009-14. 8. Kovacic KJ. Using common-sense knowledge for computer menu planning [PhD dissertation]. Cleveland, Ohio: Case Western Reserve University; 1995. 9. Khan AS, Hoffmann A. An advanced artificial intelligence tool for menu design. Nutr Health 2003; 17 (1): 43-53. 10. Khan AS, Hoffmann A. Building a case-based diet recommendation system without a knowledge engineer.Artif Intell Med. 2003;27 (2): 155-79. 11. Noah S, Abdullah S, Shahar S, Abdul-Hamid H, Khairudin N, Yusoff M, Ghazali R, Mohd-Yusoff N, Shafii N, Abdul-Manaf Z. DietPal: A WebBased Dietary Menu-Generating and Management System. Journal of Medical Internet Research 2004; 6 (1): e4. 12. Dollahite J, Franklin D, McNew R. Problems encountered in meeting the Recommended Dietary Allowances for menus designed according to the Dietary Guidelines for Americans. J Am Diet Assoc 1995; 95 (3): 341-4, 347; quiz 345-6. 13. Food and Nutrition Board (FNB), Institute of Medicine (IOM): Dietary Reference Intakes: Applications in Dietary Planning. Washington, DC: National Academy Press; 2003
Methods Inf Med 5/2005
664 Gaál et al.
14. Food and Nutrition Board (FNB), Institute of Medicine (IOM): Dietary Reference Intakes for Energy, Carbohydrate, Fiber, Fat, Fatty Acids, Cholesterol, Protein, and Amino Acids (Macronutrients). Washington,DC: National Academy Press; 2002 15. Bucolo M, Fortuna L, Frasca M, La Rosa M, Virzi MC, Shannahoff-Khalsa D. A nonlinear circuit architecture for magnetoencephalographic signal analysis.Methods Inf Med. 2004; 43 (1): 89-93. 16. Laurikkala J, Juhola M, Lammi S, Viikki K. Comparison of genetic algorithms and other classification methods in the diagnosis of female urinary incontinence. Methods Inf Med 1999; 38 (2): 125-31.
Methods Inf Med 5/2005
17. Pena-Reyes CA, Sipper M. Evolutionary computation in medicine: an overview.Artificial Intelligence in Medicine 2000; 19: 1-23. 18. Heckerling PS, Gerber BS, Tape TG, Wigton RS. Selection of Predictor Variables for Pneumonia Using Neural Networks and Genetic Algorithms. Methods Inf Med 2005; 44: 89-97. 19. Coello Coello CA.A comprehensive survey of evolutionary-based multiobjective optimization techniques, Int J Knowledge Inform Syst 1999; 1: 269-309. 20. Multi-level Multi-objective Genetic Algorithm Using Entropy to Preserve Diversity. EMO 2003, LNCS 2632, 2003. pp 148-61.
21. The M.I.T. GALib C++ Library of Genetic Algorithm Components at http://lancet.mit.edu/ga/ [verified April 11, 2005]
Correspondence to: Balázs Gaál Department of Information Systems University of Veszprém Egyetem u. 10 8201 Veszprém Hungary E-mail:
[email protected]
en17
Original Article
Personalized Dietary Counseling System Using Harmony Rules in Tele-Care István Vassányi1 , István Kósa2 , Balázs Pintér1 , Balázs Gaál1 1
University of Pannonia, Medical Informatics RD Center, Veszprém, Hungary
2
Cardiac Rehabilitation Institute of Military Hospital, Balatonfüred, Hungary
Abstract Introduction: Lifestyle assessment, especially nutrition counseling may have a great impact on the health state of everybody. The MenuGene expert system provides services for the logging and assessment of nutrition and physical activity. Aim: This paper focuses on how the harmony of a dietary log can be analyzed using harmony rules. Such an assessment can be used to assess a log as well as drive an evolutionary search process that constructs a personalized menu. Methods: Expert knowledge is formalized in two parts, sets of foods and dishes and rules that fire at a pre-defined pattern of sets. To tackle the large rule search space, a simple conflict resolution strategy is used.
Results: We implemented several hierarchies of sets to support the definition of rules, and also an Android based lifestyle assistant application. We validated the completeness of the dietary database in a survey. Conclusion: The system may prove very useful for a real improvement of the quality of life and general health state especially for patients with chronic diseases like diabetes. We plan to conduct clinical trials to prove this early next year.
Keywords Dietary menu planning, harmony rules, nutrition counseling
Correspondence to: István Vassányi University of Pannonia, Dept. of Electrical Engineering and Information Systems Address: H-8200 Veszprém, Egyetem u. 10, Hungary
EJBI 2014; 10(2):17–22 received: July 27, 2013 accepted: March 10, 2014 published: March 31, 2014
E–mail:
[email protected]
1
Introduction
developed a linear programming method for menu optimization [1], while Eckstein used random search to satisfy Nutrition has severe impact on the probability of numerical nutrient constraints [2]. Later more advanced cardio-vascular ant other diseases, so bringing tailored ad- artificial intelligence methods were developed using Casevice to the general population on nutrition and lifestyle, Based Reasoning (CBR) or Rule-Based Reasoning (RBR) including physical activity can improve life expectancy. or by combining the two methods with other techniques An important task of personalized lifestyle counseling is [3]. A web based system called DietPal has been built in dietary menu planning and analysis. The paper describes Malaysia that models the workflow of dietetic experts in the architecture and results produced by an automated order to support their work [4]. There are solutions for dietary menu generator MenuGene, applied now in a dia- parts of a nutrition counseling expert system, but so far betes home monitoring project. Our solution is intended there were no complete solutions published and validated. Our system implements all aspects of this area in a user to support, not to substitute the human dietary expert. Computer-aided menu planning and analysis is a tradi- friendly and effective way. For the first problem i.e. search satisfying numeritionally hard problem since it is characterized by i) a very large search space and ii) hard-to-formalize expert dietary cal constraints, we apply multi-level, multi-objective geknowledge on the harmony assessment of a menu. Human netic algorithms that calculate the fitness of candidate experts probably build better meal plans than comput- solutions using personalized target values of various nuers even now, although research on computerized meth- trients. The objectives for the menu planning process are ods has been ongoing since the 1960’s. In 1964 Balintfy obtained from personal medical data, entered manually c
2014 EuroMISE s.r.o.
EJBI – Volume 10 (2014), Issue 2
en18
Vassanyi et al. – Personalized Dietary Counseling System Using Harmony Rules in Tele-Care
Figure 1: Work flow of the dietary log assessment process.
or measured by sensors of the tele-care system. Another source of information is the aim of the patient like “losing weight” and the user’s daily dietary log (essentially a smart phone application) which can be analyzed for food composition and completeness. Then we apply general nutritional guidelines to compute the personalized numerical constraints at different levels. An example constraint is the daily minimum, optimum and maximum value for carbohydrate content. For the details of the Genetic Engine, see [5]. A similar approach based on evolutionary computing is reported in [6]. In this paper, however, we will focus more on the second problem, i.e. on the representation of the dietary expert knowledge.
2
Methods
For the assessment of harmony, we defined dietary concept sets e.g. ‘fruits with a high glycemic index’, and we use a simple mechanism for enforcing harmony rules over them. The rules are used to score candidate solutions e.g. a daily menu, and the score is combined with the numerical fitness of the solutions at different levels (dish, meal, daily menu). Rules assign a positive or negative score to a co-occurrence pattern of two or more sets. Positive scores mean recommended patterns, like “muesli and any drink for breakfast”, while negative scores mean detrimental combinations, like “beer and water-melon in the same meal”. The fitness value of a candidate solution is multiplied by the factor that appears in the consequence part of the applied rule, so “negative” scores mean fitness factors in the range of 0..1. Fig. 1 shows an overview on the dietary log assessment process. We support harmony rules on the meal, daily and weekly level. The condition part of a rule may contain one or more sets in an AND construct. For a meal level rule, this means that the rule will be fired if the meal EJBI – Volume 10 (2014), Issue 2
contains all of these sets, or dishes/foods in their subsets. For example, the set “fruits” contains the set “fruits with a high glycemic index” so if the rule contains “fruit” in the condition part, then all elements of the set “fruits with a high glycemic index” will match, such as ripe banana. Due to this generalized method of rule matching, several (contradicting) rules may match a meal or a day, so conflict resolution is inevitable. Our strategy is based on the following principles. • Rules with a more elements in their condition part are preferred over simpler rules because they match the actual meal/day/week more precisely • Concrete rules with less general sets in their condition parts are preferred over more general rules, for the same reason. For the case of ripe banana, the set “fruits with a high glycemic index” is a stronger reference, than “fruits”. • If two or more rules have the same complexity or generality, we apply the stricter one. A rule is stricter if in the consequence part it contains a lower fitness factor in the case of a negative rule or a higher fitness factor in the case of a positive rule. When assessing a weekly menu, or alternative weekly menus, the harmony factors of a lower level can be expressed as a vector. The computational complexity of the calculation of this vector is exponential in time in the function of the number of the slots (n), because each subset of the objects associated with slots should get evaluated according to harmony. The harmony of each subset will be expressed through the harmony payloads. Not including the empty set, the number of subsets is 2n ? 1, and this many checks are needed to calculate the value of each harmony payload. This makes the proof of the decision problem ’whether the assignment’s payloads are within the constraints’ verifiable in exponential time. For NP c
2014 EuroMISE s.r.o.
en19
Vassanyi et al. – Personalized Dietary Counseling System Using Harmony Rules in Tele-Care
Figure 2: Logging example screen shots of the Lavinia lifestyle mirror application.
complexity the proof would have to be verifiable in polynomial time by a deterministic Turing machine. Therefore, the introduction of harmony makes the decision problem harder than NP, a good candidate for evolutionary solving methods. A similar extension of the Mixed Integer Linear Problem with first-order logic is presented in [7]. Due to the large number of concrete or general, possibly applicable rules for every single meal, we store the sets and the rules in a pre-processed form that can be searched easily. The final goal of the development is to ensure real time harmony assessment of user-logged of system generated menus.
3
Results
The Hungarian version of the MenuGene data base currently stores 9500 food items along with their nutrient contents and 1373 dishes composed from the foods, but on the user interface we show only the most important 299 dishes and 360 foods, organized in 195 sets, to simplify the search. In a recent survey, we evaluated the completeness c
2014 EuroMISE s.r.o.
of our food and dish data base by manually entering 110 days of menu planned by human dietetians in a Hungarian rehabilitation hospital. The results showed that the database contained ca. 74% of the total 194 dishes that occurred in the menus, but most of the missing dishes occurred rather rarely, so if we consider the dish coverage by food diary item, the coverage ratio is 84%. This means that we could support the most common items. For the foods, these ratios are 84% and 96.5%, respectively. For this trial, we used the Lavinia lifestyle mirror application with an android based GUI (Figs. 2 and 3). This application uses the central MenuGene database and related services in a service oriented open architecture. A distinguishing feature of Lavinia compared to other lifestyle logger applications e.g. [8, 9, 10] is its integration with physiological sensors and its enhanced medial intelligence. The set hierarchy was developed manually by our dietitian expert such that it should be in line with recent professional recommendations and also support the easy construction of harmony rules. We now have a total of 1409 sets in several hierarchical structures or ontologies. EJBI – Volume 10 (2014), Issue 2
en20
Vassanyi et al. – Personalized Dietary Counseling System Using Harmony Rules in Tele-Care
Figure 3: Evaulation example screen shots of the Lavinia lifestyle mirror application.
The reason of using multiple ontologies is the need to support different grouping and classification aspects of foods and dishes. The depth of the deepest hierarchy is 7 levels (Fig. 4).
We implemented the concept of the harmony rules described above as a plug-in in the MenuGene expert GUI. Figure 5 shows a positive rule that defines the possible sets of a 2 component standard breakfasts, like “a food from the Muesli/cereal set and a drink from the set of breakfast drinks”. Sets are also used to generate warning messages for the users with specific needs or illnesses. The MenuGene system currently supports 21 chronic illnesses like Crohn disease or diabetes, and various combinations of these provided the number of concurrent illnesses does not exceed 7. The interplay of illnesses is considered for the personalized RDA limits and additionally, users with certain illnesses receive a warning when logging certain types of food. For example, a user with diabetes will receive a warning when logging coke, even if she is within her RDA. Figure 6 shows the expert GUI for defining such message rules. Another module that will allow the trace back of the rule firing process after the assessment of a menu is also in preparation.
4
Figure 4: Part of the set hierarchy.
EJBI – Volume 10 (2014), Issue 2
Conclusion
The paper presented the MenuGene lifestyle assessment expert system with an emphasis on the rules expressing dietary harmony. A simple rule model, based c
2014 EuroMISE s.r.o.
en21
Vassanyi et al. – Personalized Dietary Counseling System Using Harmony Rules in Tele-Care
Figure 5: A breakfast structure based on sets.
Figure 6: Sets used for generating warning messages on the expert GUI.
on a hierarchy of sets is used to represent domain knowl- ences”, project number: TÁMOP-4.2.2.A-11/1/KONVedge. The system has an android-based user interface that 2012-0073. supports logging and automated lifestyle assessment. We have already performed some trials to check the completeness of the data base and the usability of the user interface. References We also plan to execute clinical trials early next year to [1] Balintfy, J. L. (1964 April) ‘Menu Planning by Computer’ examine whether the use of such services really improves Communications of the ACM, vol. 7, no. 4, pp. 255-259. patients’ everyday life and general state of health, especially in such areas as diabetes and chronic kidney disease. [2] Eckstein EF. (1967 Dec) ‘Menu planning by computer: the random approach’. J Am Diet Assoc;51(6):529-533.
Acknowledgments The work presented was supported by the European Union and co-funded by the European Social Fund, project title: “Telemedicine-focused research activities in the field of Mathematics, Informatics and Medical Scic
2014 EuroMISE s.r.o.
[3] C.R. Marling, G.J. Petot, L.S. Sterling (1999) ‘Integrating Case-Based and Rule-Based Reasoning to Meet Multiple Design Constraints’. Computational Intelligence, Volume 15, Number 3 [4] Noah S, Abdullah S, Shahar S, et al. (2004) ‘A Web-Based Dietary Menu-Generating and Management System’. J Med Internet Res 2004;6(1):e4
EJBI – Volume 10 (2014), Issue 2
en22
Vassanyi et al. – Personalized Dietary Counseling System Using Harmony Rules in Tele-Care
[5] B. Gaál, I. Vassányi, G. Kozmann (2005) ‘A Novel Artificial Intelli-gence Method for Weekly Dietary Menu Planning’. Methods Inf Med 2005 (Vol. 44): Issue 5 2005, pp 655-664
[8] V. Roesler, C. Iochpe, A. Bordignon, L. Tizatto. Mobilicare: AHealth Monitoring System for Chronic Patients. proc. Med-e Tel 2013, Luxembourg, pp. 74-78.
[6] Seljak, B.K. (2006) ‘Dietary Menu Planning Using an Evolutionary Method’. Proc. Int. Conf. on Intelligent Engineering Systems, London, UK, 26-28 June 2006, pp. 108-113.
[9] Diet Assistant. https://play.google.com/store/apps tails?id=com.aportela.diets.view
[7] G. J. Gordon, S. A. Hong, and M. Dud´?k, “First-order mixed integer linear pro-gramming,” in Proceedings of the 25th Conference on Uncertainty in Artificial Intelligence (UAI), 2009.
EJBI – Volume 10 (2014), Issue 2
/de-
[10] My Diet Diary Calorie Counter. https://play.google.com /store/apps/details?id=org.medhelp.mydiet
c
2014 EuroMISE s.r.o.
Journal of Biomedical Informatics 43 (2010) 469–484
Contents lists available at ScienceDirect
Journal of Biomedical Informatics journal homepage: www.elsevier.com/locate/yjbin
An ontological modeling approach to cerebrovascular disease studies: The NEUROWEB case q Gianluca Colombo e,1, Daniele Merico a,1, Giorgio Boncoraglio f, Flavio De Paoli e, John Ellul b, Giuseppe Frisoni e, Zoltan Nagy d, Aad van der Lugt g, István Vassányi c, Marco Antoniotti e,* a Terrence Donnelly Centre for Cellular and Biomolecular Research (CCBR)/Banting and Best Department of Medical Research, University of Toronto, 160 College Street, Toronto, Ontario, Canada M5S 3E1 b Department of Neurology, University of Patras, Greece c Department of Information Systems, University of Pannonia, Veszprem, Egyetem u. 10, H-8200, Hungary d Department of Neurology, Sommelweiss University, 1088 Budapest VIII, Balassa u. 6, Hungary e Dipartimento di Informatica, Sistemistica e Comunicazione (DISCo), Università degli Studi di Milano Bicocca, U14 Viale Sarca 336, I-20126 Milan, Italy f Istituto Neurologico Carlo Besta, Via Celoria 11, I-20133 Milan, Italy g Department of Radiology, Erasmus MC, University Medical Center, ’s-Gravendijkwal 230, 3015 CE Rotterdam, The Netherlands
a r t i c l e
i n f o
Article history: Received 12 October 2008 Available online 13 January 2010 Keywords: Biomedical ontologies Data integration Clinical phenotypes Association studies
a b s t r a c t The NEUROWEB project supports cerebrovascular researchers’ association studies, intended as the search for statistical correlations between a feature (e.g., a genotype) and a phenotype. In this project the phenotype refers to the patients’ pathological state, and thus it is formulated on the basis of the clinical data collected during the diagnostic activity. In order to enhance the statistical robustness of the association inquiries, the project involves four European Union clinical institutions. Each institution provides its proprietary repository, storing patients’ data. Although all sites comply with common diagnostic guidelines, they also adopt specific protocols, resulting in partially discrepant repository contents. Therefore, in order to effectively exploit NEUROWEB data for association studies, it is necessary to provide a framework for the phenotype formulation, grounded on the clinical repository content which explicitly addresses the inherent integration problem. To that end, we developed an ontological model for cerebrovascular phenotypes, the NEUROWEB Reference Ontology, composed of three layers. The top-layer (Top Phenotypes) is an expert-based cerebrovascular disease taxonomy. The middle-layer deconstructs the Top Phenotypes into more elementary phenotypes (Low Phenotypes) and general-use medical concepts such as anatomical parts and topological concepts. The bottom-layer (Core Data Set, or CDS) comprises the clinical indicators required for cerebrovascular disorder diagnosis. Low Phenotypes are connected to the bottom-layer (CDS) by specifying what combination of CDS values is required for their existence. Finally, CDS elements are mapped to the local repositories of clinical data. The NEUROWEB system exploits the Reference Ontology to query the different repositories and to retrieve patients characterized by a common phenotype. 2009 Elsevier Inc. All rights reserved.
1. Background The NEUROWEB project aims to support association studies in the field of cerebrovascular disease, by providing a framework for clinical data integration and exchange. In general, the purpose of association studies is to find statistical relationships between a (set of) feature(s) and a phenotype, which is a composite observa-
q This work has been supported by the EC FP6 project NEUROWEB Grant 518513 and partially by the EC Marie Curie IRG Grant 031140. * Corresponding author. E-mail addresses:
[email protected],
[email protected] (M. Antoniotti). 1 These authors contributed equally to this paper.
1532-0464/$ - see front matter 2009 Elsevier Inc. All rights reserved. doi:10.1016/j.jbi.2009.12.005
ble state of an individual organism. NEUROWEB phenotypes specifically correspond to the pathological state of the patients, formulated according to clinical data collected during the diagnostic activity. In addition, the project is specifically committed, although not restricted, to provide support for exploring genotype–phenotype relations [1,2]. In order to achieve statistical robustness in these studies, large patient cohorts are required. To accomplish this goal, the NEUROWEB project involves four clinical sites from different EU-member countries, which are recognized excellence centers in the field of cerebrovascular diseases, with a particular focus on ischemic stroke. Although all sites comply with international guidelines, they have developed specific competencies in different areas of stroke diagnosis, treatment and stroke research (such as imaging,
470
G. Colombo et al. / Journal of Biomedical Informatics 43 (2010) 469–484
biochemical essays, etc.). This situation of compliance to general requirements, though with specific skills and research commitments, is mirrored by the content of their clinical repositories, which were independently developed to store the patients’ profiles gathered during the diagnostic activity. The size of each data repository varies from 500 to 1500 patient records. The exploitation of NEUROWEB local repositories for association studies presents two major modeling aspects (cf. [3–5]): (1) developing an IT infrastructure enabling the access to different technological platforms for data storage (interoperability); (2) resolving the semantic misalignments among locally-defined clinical indicators, yet preserving the methodological coherence and consistency of phenotype formulation. Specifically, the misalignment among clinical indicators does not rest on a purely linguistic or terminological basis, as displayed by the following cases. (1) The same type of exam can be performed using different diagnostic methodologies or technology (e.g., CTA scan vs. MRI scan), with different implications for diagnostic reliability. (2) Different findings can be derived from the raw results of the same type of exam; as a consequence, the same clinical indicator may have different values in different sites (e.g., Brachiocephalic artery lesion, derived from CTA scan results). (3) Different scale of granularity, intended as level of detail concerning the patient’s state assessment (e.g., ICA–CCA Stenosis Present/Absent vs. ICA–CCA Stenosis Left Present / Absent). (4) Different criteria can be applied to assign a stroke classification label (e.g., Atherosclerotic Ischemic Stroke) by combining different clinical indicators, or by setting different criteria (e.g., ranges, thresholds, acceptable values, etc.) on the same indicators. Many state-of-the-art solutions for data integration (for health care as well as other domains) revolve around the idea of database schema matching, i.e., establishing relations between elements from local databases [6,7]. Database schema matching can be supported by semantic models, specifying the meaning of concepts behind database elements. Such models can range in complexity from simple graphs to semantic networks and domain ontologies [8,9]. An important distinction can be set between systems relying on a single, global conceptual schema or local, independent conceptual schemas. The two solutions have specific advantages and disadvantages: the global schema solution requires a constant update of the semantic model whenever local databases change, whereas local schemas can be hard to reconcile unless they are already based on shared concepts or terminologies. Of course, there are already several systems exploiting terminological resources, ‘‘semantic mediators” or ontological knowledge models for data integration and retrieval; a few examples are [10–13]. Also, several Ontology-based (OB)-Systems supporting highthroughput processing of biological and clinical data have appeared [14–21]. These systems are traditionally aimed at gathering genotypic information associated to patients [21] through the adoption of bio-ontologies. The genotypic information gathered is then used either to guide data selection and knowledge discovery processes [18], or for biomedical data integration, extraction and mining [14,17,20]. A unified panorama of available bio-ontologies is offered by the National Center for Biomedical Ontology (NCBO) Bioportal [19].
With respect to existing OB-Systems supporting phenotype– genotype scientific inquiries, NEUROWEB primarily copes with the representation of the clinical rather than biological knowledge involved in association studies. Moreover, case-control studies have to be exhibited [22] in order to strengthen an association hypothesis between a given polymorphism and phenotypic profiles of cerebrovascular diseases [23,24]. Such investigation requires both a robust clinical knowledge modeling (i.e., clinical phenotype ontology) and the definition of ad hoc system modules that guarantee the methodological coherence in data mapping, data extraction and data collection (i.e., IT-facilities described in the following sections). No state-of-the art systems nor state-ofthe-art ontologies on clinical phenotypes modeling (as discussed in [25]) to support clinical based testings of hypothesized phenotype–genotype associations were available at the time the NEUROWEB project began, nor, to the best of our knowledge have been developed in the meantime. Given this background, the NEUROWEB project set forth to provide its own form of integration. In the case of NEUROWEB Project, semantic reconciliation goes a step further than straightforward database schema matching. In fact, the NEUROWEB system was designed to offer clinicians, and also biologists, the capability of performing collaborative research through access of a network of data repositories. Each repository could be queried either using unified, higher-level concepts referring to common-use cerebrovascular phenotypes, or using new, user-defined phenotypes, assembled from elementary phenotype units. To achieve this goal, we developed an ontological model [26–30] for cerebrovascular phenotypes – the NEUROWEB Reference Ontology – that accurately captures the core diagnostic/classificatory knowledge of clinicians. The development of the Reference Ontology required very intensive knowledge acquisition and knowledge structuring activities, leading to a progressive refinement of the semantic model, as described in Section 2. In the final model, top-level diagnostic classes are deconstructed into more elementary phenotypes and medical concepts. Every phenotype is associated to the specific combination of clinical indicators required for its occurrence. The content of local databases is mapped to the Reference Ontology locally; the current version relies on direct mapping and simple rules, but the future adoption of local ontologies, to capture local features in a more expressive way, is supported as well. Specifically, mapping local database elements to the ontology elements enables the reconciliation of granularity discrepancies between locally-defined clinical indicators that would be hard to manage using direct mappings between such elements. The adoption of a Reference Ontology grounded on expert knowledge grants the respect of methodological consistency, which is an essential requirement for association studies. A major decision to be taken was to choose between the development of a specific Reference Ontology and the adoption of an existing one [31]. The second solution would apparently offer superior advantages, by granting inter-operability with external resources, and by facilitating the involvement of new partners in case they are already complying with an existing ontology. However, there are crucial problems undermining this solution. Indeed, no publicly-available medical ontology is committed to the representation of clinical findings and phenotypes.2 The phenotype ontologies developed within the biological community are oriented to high-throughput genetic experiments in model organisms [29,33,34], and hence are not suitable for clinical applications. From a pragmatic perspective one may argue that even if an optimal solution is not available, the reuse of an existing general-purpose medical ontology may still offer significant advantages. As a case in point,
2
This is a very active research area; see, for example [32]
G. Colombo et al. / Journal of Biomedical Informatics 43 (2010) 469–484
we did consider SNOMED-CT as a possible candidate, but decided against it because of a number of actual shortcomings that would be encountered by adopting it (see also [35–37]). Considering the stroke type taxonomy used by NEUROWEB clinicians, most of the concepts are either missing from SNOMED-CT, or they are formulated in an unsuitable way. For instance, the definition of SNOMED-CT Atherosclerotic Occlusive Disease clearly implies an atherosclerotic etiology, but not the specific features of stroke, which are part of the NEUROWEB Ischemic Stroke definition. In addition, several clinical indicators required for stroke diagnosis and present in NEUROWEB clinical repositories are missing from SNOMED-CT (e.g., relevant scan lesion). SNOMED-CT offers qualitative scales for clinical findings but does not provide quantitative criteria to assign them (e.g., no stenosis percentage ranges are associated to the previous scale), nor does it resolve inter-dependencies among different indicators. Similar shortcomings are displayed by other medical resources, proving that these problems are not specific of SNOMED-CT only. For instance, the Disease Ontology [38] is a general-purpose classification of pathologies, describing biological-samples within genetic data banks (developed for the NUgene project). The Disease Ontology includes concepts which have terminological correspondence to NEUROWEB phenotypes and clinical indicators (e.g., Stroke, Atherosclerosis, Subarachnoid hemorrhage, Cerebral embolism, Cerebral thrombosis, Occlusion and stenosis of carotid artery, etc.). However: (1) they do not follow the taxonomy adopted by the cerebrovascular clinicians who are part of the NEUROWEB network; (2) no criteria are provided to assign them on the basis of clinical data; (3) the concepts are organized by adopting only is-a, part-of, inverse-of, union-of and disjoint-from relations, thus lacking any specification of causality or diagnostic evidence. As a whole, there is a wealth of knowledge, relevant to the aims of the NEUROWEB system, which could not be properly conveyed by this ontology. Consequently, it was decided to develop a new Reference Ontology committed to accurately represent the diagnostic knowledge of the NEUROWEB community. Nonetheless, NEUROWEB concepts were mapped to external terminologies to support keyword-based searches in external resources. The value and validity of the diagnostic knowledge encoded by the NEUROWEB Reference Ontology is not restricted to the NEUROWEB Consortium, as the stroke type taxonomy adopted largely follows the TOAST classification guideline. This resource was developed by cerebrovascular experts, and is regarded as a reliable standard by the international cerebrovascular community [39–41]. 2. The Development of the NEUROWEB Reference Ontology The development of the NEUROWEB Reference Ontology proceeded iteratively over a number of phases. In the narrative of Section 2.1, we identify three major phases where design decisions where made and provide a rationale for the choices made. 2.1. Initial models 2.1.1. Basic level definition The modeling activity started by identifying a minimal set of clinical indicators (Core Data Set, in brief CDS), mapped to the majority of the local repositories and of primary importance for stroke diagnosis. In a sense, the CDS is the equivalent of one of
471
the numerous minimal information specifications now coordinated, e.g., in the MIBBI project [42]. The definition of the CDS was characterized by close collaboration with clinicians, and several cycles of refinement at round table meetings. The most straightforward solution for clinical data integration consists in mapping the content of local repositories to the CDS. In this scenario, the user formulates queries on the CDS indicators, which are then translated on local databases. This solution is problematic, as there are often granularity discrepancies between local indicators, preventing a direct mapping to CDS elements (e.g., scan lesion = yes vs. scan lesion side = LEFT). In addition, some CDS indicators are not atomic, meaning that they implicitly refer to other entities of the CDS. For instance, relevant scan lesion does not refer to a single exam result; according to diagnostic knowledge, a lesion is relevant only in the presence of a co-axial scan lesion (i.e., the evidence of some brain tissue damage) and stenosis (i.e., the evidence of a partial occlusion in a brain-afferent artery). It follows that relevant scan lesion = yes requires a set of constraints on several atomic entities of the CDS, namely: scan lesion = yes, stenosis degree > 50%, side of the scan lesion = RIGHT (LEFT), and side of the stenosis = RIGHT (LEFT). Stroke classification labels (e.g., TOAST Atherosclerotic Ischemic Stroke Evident) are a specific type of non-atomic CDS indicators. They characterize the patient’s state in a comprehensive way, implicitly encompassing many other indicators. These classification labels could be used to directly retrieve patients for association studies. However, since they were manually assigned by clinicians, errors and methodological inconsistencies are possible. In particular, it is possible that different clinicians or different sites used different criteria for their assignment. For this reason, using the stroke classification labels would be a satisfactory solution under the mere perspective of interoperability, but methodological coherence and consistency, which are crucial for rigorous association studies, may not be ensured. In conclusion, the CDS alone is insufficient and a richer model is required, utilizing the CDS as the groundwork. 2.1.2. Two-layers solution We next developed a model composed of two layers, where the top-layer is a taxonomy of phenotypes (Top Phenotypes), each connected to the CDS entities (the bottom-layer) via a definition formula. The formula is structured as a conjunction/disjunction of criteria on CDS indicators, expressed as equality/inequality bounds, or quantitative ranges to be satisfied. For instance: (Blood pressure > 50) AND (Anterior cerebral artery lesion = yes). The model above was used for an extensive knowledge acquisition activity with clinicians. It offers the advantages of being simple to understand, and it includes entities already familiar to the clinicians. Using the CDS entities to formulate Top Phenotypes also provided valuable feedback for the refinement of the content and structure of the CDS. 2.1.3. Three-layer solution The major weakness of the two-layer model is the absence of relations deconstructing the stroke types into more elementary phenotypes and medical concepts of general use (e.g., co-occurring pathologies such as diabetes and obesity, anatomical parts, topological concepts). This additional array of entities is necessary to support important functionalities.
472
G. Colombo et al. / Journal of Biomedical Informatics 43 (2010) 469–484
Reference Ontology Ischemic Stroke
Top Phenotypes (stroke taxonomy) Phenotype Ontology Topology & Anatomy
Low Phenotypes (building blocks)
Indicator Value
Mapping Rule
Core Data-Set (clinical indicators)
NEUROWEB to Local Mapping
Local Clinical Repositories Fig. 1. The diagram displays the overall architecture of the NEUROWEB Reference Ontology.
Simplify the handling of the phenotype formulas; in fact, subunits of the full definition formula can be associated to certain aspects of the pathological state, of general validity even in a different context than the stroke specialty.3 Establish a mapping to medical ontologies and terminologies (e.g., SNOMED-CT, MeSH [43,44]), in order to support document retrieval, and other searches on resources outside the NEUROWEB consortium. Support the integration of bioinformatics resources for genotype–phenotype association (e.g., the Human Gene Mutation Database – HGMD [45,46]) and gene function (most notably from the Gene Ontology [47]). These problems can be overcome by adopting an ontological framework for phenotype formulation and introducing an additional layer (Low Phenotypes) to deconstruct the Top Phenotypes into more elementary concepts. Henceforth, the resulting model is composed of three layers, and it will be fully described in Section 2.2. 2.2. The final ontology architecture The NEUROWEB Reference Ontology is composed of three layers (Fig. 1). Top Phenotypes: taxonomy of stroke types. Low Phenotypes: anatomical parts, topological relations, diseases and other phenotypes. Core Data Set (CDS): unified clinical indicators. The Top and Low Phenotype are grouped into the Phenotype Ontology, which, alongside the CDS, is part of what we refer to as the NEUROWEB Reference Ontology. The CDS (bottom level) is the set of minimal-granularity clinical indicators required for stroke diagnosis. The Phenotype Ontology is composed of Top Phenotype and Low Phenotype layers. The Top Phenotype layer is a taxonomy 3 For instance, the Top Phenotype Atherosclerothic Ischemic Stroke can be decomposed into two parts: the Ischemic Stroke (a cerebrovascular accident), and its durative etiological factor Atherosclerosis (a vessel disease). Identifying these two parts is useful, as the Ischemic Stroke unit is also a part of Cardioembolic Ischemic Stroke, another Top Phenotype.
of stroke types, connected to the Low Phenotype layer by relations such as causality and existence of diagnostic evidence; the deconstruction of Top Phenotypes into low Phenotypes provides an explicit model for the inherent classification criteria underlying the Top Phenotype taxonomy. Finally, the Low Phenotypes are connected to the CDS entities and the corresponding combination of values required for the phenotype to occur. Mapping from NEUROWEB Reference Ontology elements to local repositories occurs primarily at the CDS level. However, mapping to Low Phenotypes is also possible; this solution ensures the handling of the aforementioned granularity discrepancies between local repositories (cf. Section 2.1), as the phenotype ontology provides progressively more general concepts than the CDS. For more details on the mapping solution, please refer to Section 3.2.
2.2.1. The Core Data Set In the final, three-layer model, the CDS was restricted to minimal-granularity clinical indicators. In fact, non-atomic concepts are handled within the Phenotype Ontology (e.g., stroke classifications are represented by Top Phenotypes). For this reason, mapping to local repositories is not restricted to the CDS layer. The CDS entities are organized into categories and sub-categories, according to the different types of examinations (see Fig. 2). In general, the values of CDS entities can be quantitative (e.g., Age, Hemoglobin on admission g/dL), Boolean (e.g., Current use of alpha blockers: yes, no), or categorical (e.g., Cognitive Function: normal, mildly impaired, confused; Gender: male, female).
2.2.2. The Phenotype Ontology The upper-layer of the Phenotype Ontology is populated by the Top Phenotypes. These entities represent the main classes of pathological states typically diagnosed by clinicians. They are inter-related by is-a relations, thus forming a taxonomy of stroke types. The root of the taxonomy is Ischemic Stroke, whose children are Atherosclerotic Stroke, Cardioembolic Stroke and Lacunar Stroke; these stroke types constitute the three main etiological groups, as they are identified in the TOAST classification. Each of them is then divided into Evident, Probable and Possible. Deeper levels of phenotype specification (e.g., according the anatomical location of the vascular lesion) lead to the addition of children to these stroke subtypes.
G. Colombo et al. / Journal of Biomedical Informatics 43 (2010) 469–484
Main Categories
Sub-Categories
Example Entities Gender Age Cholesterol on Admission Facial Palsy
Personal Identifying Data Clinical Data Brain Imaging Studies
473
CT (Computed Tomography)
Location of the Lesion
MRI (Magnetic Resonance Imaging)
Heart Studies
Vessel Studies
Holter
Paroxysmal Atrial Fibrillation
ECG (Electrocardiogram)
Rhythm on Admission
Transesophageal-echocardiogram
Right-left Shunt
Duplex
Internal Carotid Artery Stenosis
CTA (Computed Tomography Angiography) MRA (Magnetic Resonance Angiography)
Carotid Bifurcation Stenosis
DSA (Digital Subtraction Angiography) Laboratory Studies
Follow-up Information
Medication at Follow-up
Routine Blood Tests
Hemoglobin on Admission
Coagulation Studies
Fibrinogen
Follow-up Information Month 1 Follow-up Information Month 3 Follow-up Information Month 6 Follow-up Information Month 12 Medication at Follow-up Month 1 Medication at Follow-up Month 3 Medication at Follow-up Month 6 Medication at Follow-up Month 12
Classification
Cognitive Function
Current Use of Alpha-blockers TOAST - Classical TOAST - Extended ICD-9CM OCSP
Fig. 2. The table displays the categories and sub-categories used to organize the CDS entities. An example of CDS entity is provided for each partition. The classification layer is shaded, as these indicators are conveyed by Top Phenotypes, and were part of the CDS only in the initial models.
To define the entities and relations deconstructing the Top into Low Phenotypes,we followed the TOAST classification criteria. Etiology (Atherosclerotic, Cardioembolic, Lacunar Stroke). Confidence of the etiological assessment (Evident, Probable, Possible), depending on the strength of the diagnostic evidence for the most-probable etiology. In addition, anatomy (i.e., the location of the lesion) is not used by the TOAST, but could be suitably used to extend it, and is consistently used in the different clinical communities. These criteria are explicitly recognized by the clinicians, and are generally used in clinical medicine. It is also important to consider that diagnostic activity is always characterized by the acquisition of diagnostic evidences, enabling the reconstruction of the undergoing patho-physiological processes and structures, even if they are not directly observed. In the specific case of ischemic stroke, there is a consistent partition between the evidences for the ischemic damage (typically brain imaging displaying the damaged tissue, and the cognitive/motor impairment of the patient) and the evidences for the cause of the occlusion causing ischemia; the latter is usually a persistent or progressive state of the patient’s organism, such as Atherosclerosis, whereas the latter occurs after a chain of point-events (i.e., with a very compact timespan) eventually leading to a brain lesion (a trauma). From a biological standpoint, however, it is important to represent not only the diagnostic evidences, but also the patho-physiological processes inferred. In our modeling case, ischemic stroke is caused by the rupture of an atherosclerotic plaque, triggering the coagulation cascade, the release of a clot particle into the bloodstream (embolization) and the obstruction of a brain artery, even-
tually resulting in a brain lesion in the region deprived of the blood supply. These processes can be captured at the biomolecular level and represented in the ontology. Reflecting the above mentioned criteria, the Top Phenotypes are first decomposed into Low Phenotypes (see Fig. 3) according to etiology. We used two different causal relations: Has-Cause-Durative, Has-Cause-PointEvent. Has-Cause-Durative connects a Top Phenotype to its Durative Etiological Background (e.g., Atherosclerotic Disease), representing the long-term pathology responsible for the generation of the ischemic event. Has-Cause-PointEvent connects a Top Phenotype to its Traumatic Point Event, representing the cerebrovascular accident that occurred in the patient. As far as the current version of the ontology is concerned, this will be always IschemicEvent or any of its specifications. However, cerebrovascular accidents other than Ischemic Stroke can be represented as well (e.g., Hemorrhagic Stroke). This first group of Low Phenotypes reflects patho-physiological processes. The underlying biomolecular processes are connected via the Involves relation. For more details on how biomolecular processes are exploited by the system, please refer to Section 2.4.1. Both Durative Etiological Background and Traumatic Point Event are then connected to their diagnostic evidences (Durative or Pointevent, respectively), via the Has-Diagnostic-Evidence relation. Diagnostic evidences can be decomposed into more elementary diagnostic evidences using the same relation. An additional dimension to be taken into account is the anatomical location implied by the phenotype. Anatomical parts cannot
474
G. Colombo et al. / Journal of Biomedical Informatics 43 (2010) 469–484
Fig. 3. The diagram displays the main types of entities and relations composing the NEUROWEB Reference Ontology. The three layers of the Reference Ontology (Top Phenotypes, Low Phenotypes, CDS) correspond to the large shaded boxes in the background on the right side. Additional entity types are displayed on the left side. Colored arrows represent relations. The graphical pattern A ! B1 ! B2 represents the DL construct: A $relationX.(B1 u $relationY.B2).
be regarded as phenotypes (i.e., observable properties): they are rather physical entities bearing observable properties [29–31,33]. For this reason, we introduced new types of entities, Anatomical Parts and Topological Concepts. Diagnostic evidences are connected to Topological Concepts via the Has-Side relation, and to Anatomical Parts via the Has-Location relation. Low Phenotypes of type Diagnostic Evidence are finally deconstructed into the exams required for their assessment. The Has-Value relation allows for the formulation of validity ranges that must be satisfied by a CDS indicator, in order to elicit the occurrence of a certain phenotype; the resulting construct is connected to the pertinent Low Phenotype through the relation By-Means-Of. As an example of integration feasibility, the Phenotype Ontology can be effectively mapped to other medical ontologies, in order to support queries on external resources. At the present stage of model development we provide the mapping between SNOMEDCT terms and Reference Ontology entities, belonging to the Low Phenotypes and the Anatomical Parts modules. The mapping cannot be systematically provided for the Top Phenotypes and CDS entities, as SNOMED-CT does not offer a satisfactory coverage, for the reasons already explained in the Section 1.
Table 1 Description Logic code for the fragment of the NEUROWEB Reference Ontology Atherosclerotic Ischemic Stroke Evident (AISE).
2.3. An example of phenotype formulation
nected to the topological indication of the side via the Has-Side relation. Moderate and Severe Lesion are eventually decomposed into CDS elements and appropriate diagnostic values by the By-MeansOf and Has-Value relations.
As an example, let us have a look at the ontological definition for the Top Phenotype Atherosclerotic Ischemic Stroke Evident (AISE); see Table 1 and Fig. 4. The first axiom states (a) the existence of a relation from Atherosclerotic Ischemic Stroke Evident (AISE) to the Low Phenotype Ischemic Event via the Has-Cause-PointEvent relation, and (b) the existence of a relation from AISE to the Low Phenotype Atherosclerotic Disease via the Has-Cause-Durative relation. As an additional requirement, Atherosclerotic Disease must have Severe Stenosis as diagnostic evidence – this is a specific requirement of Atherosclerotic Ischemic Stroke when it is Evident. According to the second axiom, Ischemic Event requires the presence of the diagnostic evidence Relevant Lesion; Atherosclerotic Disease is not further decomposed just for the sake of brevity. Relevant Lesion is further decomposed into Left and Right Relevant Lesion, which consist of a Moderate or Severe Lesion con-
AISE $hasCausePointEvent.IschemicEvent u$ hasCauseDurative.(AtheroscleroticDisease u$ hasDiagnosticEvidence.SevereStenosis)
(1)
IschemicEvent $ hasDiagnosticEvidence.RelevantLesion
(2)
RelevantLesion $ hasDiagnosticEvidence.LeftRelevantLesion t$ hasDiagnosticEvidence.RightRelevantLesion
(3)
LeftRelevantLesion $ hasDiagnosticEvidence.((ModerateLesion t SevereLesion) u$ hasSide.Right) u$ hasDiagnosticEvidence.(SevereStenosis u$ hasSide. Right) ModerateLesion Lesion u$ byMeansOf(CT-From2.5to5CentimetersLesion t MRI-From2.5to5CentimetersLesion t PET-From2.5to5CentimetersLesion)
(5)
CT-From2.5to5CentimetersLesion CT u$ hasValue. 2.5-5centimeters MRI-From2.5to5CentimetersLesion MRI u$ hasValue. 2.5-5centimeters PET-From2.5to5CentimetersLesion PET u$ hasValue. 2.5-5centimeters
(6) (7) (8)
2.4. Extensions of the NEUROWEB Reference Ontology In the following we will briefly discuss some extensions of the NEUROWEB Reference Ontology under development: a terminological extension to support text-search tools, a more sophisticated treatment of the temporal dimension, and in more details the addition of a layer for the treatment of biomolecular processes. One of the primary aims of the NEUROWEB Reference Ontology is also to support the retrieval of patients characterized by a certain phenotype. Nonetheless, such an semantically articulate model can be usefully exploited also for text-mining purposes, thus support-
475
G. Colombo et al. / Journal of Biomedical Informatics 43 (2010) 469–484
Atherosclerotic Ischemic Stroke
(OR) Is-a
Is-a
Atherosclerotic Ischemic Stroke Evident Has-CausePointEvent
Is-a
Atherosclerotic Ischemic Stroke Probable
(AND)
Atherosclerotic Ischemic Stroke Possible
Has-CauseDurative
Ischemic Traumatic Event
Atherosclerosis
HasDiagnosticEvidence
Severe Stenosis
Relevant Lesion HasDiagnosticEvidence Left Relevant Lesion
Right Relevant Lesion HasDiagnosticEvidence
HasDiagnosticEvidence Severe Lesion Right
HasDiagnosticEvidence
Severe Stenosis
Moderate Lesion
Right
Right
By-Means-Of
Has-Side
By-Means-Of
By-Means-Of CT.Presence
CT.Lesion.Size
CT.Lesion.Side
Yes
2-5cm
Right
Has-Value
Fig. 4. A visual diagram for the AISE DL formulation. Big boxes represent entities, arrows represent relations. Small gray boxes (transitions) are used to convey AND/OR logical connections: AND corresponds to relations using the same transition, OR corresponds to relation using independent transitions.
ing the NEUROWEB in literature searches. To effectively support the processing of texts, an ontology needs to be extended with a terminological basis, i.e., the set of written expressions corresponding to the ontology-encoded concepts [48]. A practicable solution would be to link each phenotype concept to the corresponding terms from publicly available terminological resources (e.g., SNOMED-CT, MeSH, UMLS [49], ICD-9CM [50], ICD-10 [51]), as was exemplified in Section 2.2 for SNOMED-CT. Specifically, the mapping to MeSH would be very effective in supporting Medline searches, as Medline is already indexed by MeSH terms. A case could also be made to incorporate synonym searches over WordNet [52], although such searches would mostly yield uncontestualized (in the biomedical sense) results. The terminological extension of the Reference Ontology is currently under way, and can be already used in its prototype stage. The formal representation and the computational treatment of the time dimension is a crucial topic for the data management systems [53] and the ontological knowledge representation area [54], as well as for the biomedical support systems [55] and the developing area of biomedical ontologies [56,57]. According to [55] two main research directions drive the temporal dimension issues within the biomedical and clinical information system development: (1) temporal reasoning, in order to
support inferential tasks, such as therapy planning and execution; (2) temporal data maintenance topics, pertaining to the storage and retrieval of clinical data having heterogeneous temporal dimensions. As far as the representation and reasoning issues are concerned, the temporal dimension introduced by the diagnostic activity – i.e., modeling the temporal order of the diagnostic examinations, and the process of progressive hypothesis refinement – will not be addressed, as it would be out of the NEUROWEB scope (that is, supporting association studies via patients clustering). On the contrary, the temporal dimension relevant to the NEUROWEB system concerns the formulation of phenotype on the basis of dynamic patterns, i.e., the varying behavior of clinical indicators over time. For instance, it could be valuable to define improving or worsening conditions. Such a modeling effort would require, on one hand, a systematic treatment of repeated clinical examinations over time; on the other hand, it would require an additional knowledge acquisition activity, in order to identify the criteria implicitly applied by the experts when they recognize different classes of dynamic patterns. Finally, the knowledge representation formalism could reasonably be the one proposed in [54], or an adaptation of the same.
476
G. Colombo et al. / Journal of Biomedical Informatics 43 (2010) 469–484
2.4.1. The Biomolecular extension of the Reference Ontology The NEUROWEB consortium is committed to integrate clinical and genetic databases of the participating centres, referring to patients affected by cerebrovascular diseases that were studied with high-throughput Single Nucleotide Polymorphism (SNP) [58,59] genetic profiling. The data will be analyzed to identify statistically significant genotype–phenotype associations, fostering the discovery of new diagnostic markers, disease genes and potentially therapeutic targets. Due to co-inheritance patterns [58], the analysis of SNP data typically poses a genomic mapping problem: given a certain polymorphism, which is significantly associated to the phenotype of interest, it is often hard to assess what gene mutation is actually responsible for the observed phenotype. One of the heuristics that can be used to address this problem is to combine the newer whole genome approach with the older candidate gene approach [2], selecting for genes characterized by functions compatible with the observed phenotype (e.g., cholesterol homeostasis in Atherosclerotic Ischemic Stroke) [60]. The NEUROWEB Reference Ontology was extended to biomolecular entities in order to support this goal. In addition, to provide mappings to external resources, the ontology includes Has-Reference relations connecting NEUROWEB processes to gene functions from Gene Ontology (GO), and pathways from resources such as KEGG [61–63]. Here follows a qualitative description of how the NEUROWEB system supports polymorphism mapping. The input consists of single SNP-phenotype associations. The SNP (identified by the NCBI dbSNP ID [59]) is mapped to its genetic locus. The NEUROWEB Genomic Engine retrieves all the genes mapping to the locus; the Gene Ontology (GO) functional annotation is retrieved for those genes. If any of the GO annotations are mapped to the NEUROWEB biomolecular processes, the ontology is navigated to identify one (or more) connections to Low Phenotypes via the Involved-In relation. The ontology is further navigated to search for connections between the previously identified Low Phenotypes and the phenotype of interest (i.e., the one associated to the polymorphism). The system outputs the genes and the relative biomolecular processes that are associated to a valid path. Output genes represent good candidates for the refinement of the genotype– phenotype association. The corresponding output biomolecular processes offer a plausible functional explanation of the genotype–phenotype association.4 The implementation of this functionality in the NEUROWEB system is currently under completion. The biomolecular extension of the NEUROWEB Ontology can be further exploited for the integration of NEUROWEB genotype–phenotype associations with data from publicly available resources (e.g., the NIH Genetic Association Database [64]). This functionality is currently under study; so far we have identified two major discrepancies to be addressed. Use of different criteria to identify phenotypes and diseases. Phenotypes can be referred to different levels of biological organization. 4 For instance, let’s consider the association between a given SNP and the Top Phenotype Atherosclerotic Ischemic Stroke Evident. The SNP is mapped onto an LDL Receptor gene, annotated for Cholesterol Homeostasis according to Gene Ontology. GO Cholesterol Homeostasis is connected to NEUROWEB Cholesterol Metabolism and Homeostasis, and this process is connected to the Low Phenotype Atherosclerotic Disease, which is connected to Atherosclerotic Ischemic Stroke Evident via Has-Cause-Durative. It is thus possible to establish a valid path from Atherosclerotic Ischemic Stroke (and all its offspring) to the SNP of interest, by traversing Atherosclerotic Disease and Cholesterol Metabolism and Homeostasis.
The first problem occurs because, to the best of our knowledge, there is no common ontology for clinical phenotypes working as a common exchange language, and because different communities of clinical expert may have specific needs. Thanks to the decomposition of the Top Phenotypes into the more general Low Phenotypes, the NEUROWEB Reference Ontology is already structured to minimize this problem. The second problem typically occurs when phenotype annotation refers to cellular and molecular processes or components, whereas clinical phenotypes are most often referred to the organ or system level.5 A complete solution of this problem would require a systematic ontology of biological functions and parts, spanning all the organization levels, from molecules to organs and systems. Yet, a more limited yet satisfactory solution can be achieved relating biomolecular processes to Low Phenotypes. Establishing such relations amounts to defining the molecular mechanisms of pathological manifestations (e.g., atherosclerosis, atrial fibrillation, etc.).
3. Methods 3.1. Description Logic Implementation of the NEUROWEB Reference Ontology Description logics [65–67] are a family of logic-based knowledge representation formalisms designed to represent and reason about the knowledge of an application domain in a structured and well-understood way. The basic notions in description logics are atomic concepts and atomic roles (unary and binary predicates in the terminology of first order language, respectively). In order to distinguish the function of each concept in the relation (represented by a role), the individual object that corresponds to the second argument of the role, viewed as a binary predicate, is called role filler. For instance, hasPart.Wheel is an expression which describes properties of cars having wheels, in which the individual objects belonging to the concept Wheel are fillers of the role hasPart. A specific description logic is mainly characterized by the constructors it provides to form complex concepts and roles from the atomic ones. The language we used to formalize the ontological clinical knowledge is SHOIN [65], which is an extension of the basic description logic. In order to develop the Reference Ontology computational model we have adopted the OWL DL version [68]. The editor adopted for the OWL files generation is Protégé, a well known tool developed and distributed by the Stanford University, where the Reference Ontology concepts are represented as TBox (Terminological Box) entities.
3.2. Exploiting the Reference Ontology for clinical queries The NEUROWEB phenotypes are defined in the Phenotype Ontology as axioms that cannot be directly exploited to query the local repositories, which are typically implemented with relational databases. To retrieve patient clusters on a phenotypic basis, such high-level definitions need to be typically translated and mapped into CDS entities and then into local database queries. Moreover, since a local database may include only some of the indicators represented by the CDS and possibly other elements, the actual mapping may occur at different level of the Reference Ontology. For example, a specific local database may include information about a given phenotype (e.g., presence or absence of ste5 E.g., HGMD phenotype Apolipoprotein A1 deficiency is related to an increased susceptibility to thrombogenesis and embolization in presence of ulcerated atherosclerotic plaques, and thus should be related to the NEUROWEB Atherosclerotic Ischemic Stroke.
G. Colombo et al. / Journal of Biomedical Informatics 43 (2010) 469–484
477
Fig. 5. User interface of the Ontology Mapper tool for a flat local database table. The user can select one or more local database fields (upper right) and assign them to a CDS label or a phenotype (upper left). The tool generates the syntax for the mapping to be saved as an SQL view.
nosis) that could be directly mapped to the corresponding NEUROWEB phenotype. To define NEUROWEB-to-Local (N2L) mapping, local medical experts familiar with the actual meaning, and – possibly – coding of the local repository logical structure, are requested to map the local repository elements to the Reference Ontology elements. In order to facilitate this task, a graphical interactive tool, named Ontology Mapper, has been developed as a Protégé plug-in. In most cases the mapping may occur at the CDS level since the CDS structure should be quite similar to the one of the clinical repositories. In some cases the mapping could be reduced to a linguistic translation of terms in the CDS into the local terminology, in other cases, several local data fields may be used to construct a single CDS element. However, to host new clinical partners it might be necessary to map the Reference Ontology concepts or even to develop local ontologies to express a more sophisticated relationship and exploit the full range of the NEUROWEB solution (see Fig. 5). The resulting high-level and CDS-level mappings are stored the same way at clinical servers as database views in our prototype implementation (cf., ‘‘Mapping to local entities” in Fig. 8). The CDS labels and phenotypes that appear in these views are called ‘‘supported”, while others are ‘‘unsupported” by the clinical partner. The clinical sites periodically communicate the list of their supported CDS labels and phenotypes to the center for storage and use when designing queries (cfr., ‘‘Mapping to supported entities” in Fig. 8). The support for phenotype-level mappings is useful to avoid ‘‘semantic gaps”, i.e., loss of information, in user queries. Such gaps occur when querying for a phenotype whose CDS constituents cannot be mapped to a local repository due to granularity discrepancies. Mapping at the phenotype level increases the flexibility of the system. However, the user is given the opportunity to set con-
straints on such flexibility. To that end, we use a system of weights and thresholds: the user can assign weights to specific components of the query (e.g., certain CDS elements); then each record retrieved from each data source is scored, by summing the weights of the supported fields. If this score does not satisfy a user-defined threshold, that record will not be output to the user (see Fig. 6). The process of generating queries to access a local repository requires two tasks: (1) the elements of the local repository need to be mapped into the ones in the Reference Ontology and the CDS, and (2) the NEUROWEB phenotypes need to be transformed in queries in terms of the reference ontology elements that map to the local repository. To perform the two tasks, we have developed two components that are part of the NEUROWEB architecture as illustrated in Fig. 8. (1) The Phenotype Converter that relies on the mapping information provided by every participating repository to generate tailored SQL queries in the NEUROWEB terminology. (2) The NEUROWEBto-Local (N2L) Mapper that exploits the mapping information to generate the actual queries that will be used to access the local database. The Phenotype Converter has been implemented as a Java centralized component to ensure the logic consistency over the set of performed queries, exploiting the Jena programming interface [69] to navigate the ontology and extract class names and axioms. In practice, every request is decomposed in the same way according to the current definitions included in the Reference Ontology; whenever they change only the centralized component needs to be changed without notifying the involved sites. This solution improves the awareness of the users that can be immediately notified about who is going to answer a specific query and how. The conversion process essentially navigates the references from the top level phenotype axioms to low phenotypes, and finally to conditions on CDS elements, which are the leaf nodes of the
478
G. Colombo et al. / Journal of Biomedical Informatics 43 (2010) 469–484
Fig. 6. Two example mappings to local data fields in the PL-SQL language (simplified versions of the originals). The first mapping maps the local field stroke.ctl to the phenotype Presence_Of_Carotid_Stenosis if the field contains the specified string. The second mapping creates the CDS label F.01.04.09.01.00.00 (code for Documented recurrent stroke/TIA) using the local fields stroke.nihss6ho and stroke.nihss12ho.
phenotype tree. For this navigation, the has-cause, has-evidence and by-means-of relations are used. By combining the simple conditions with the operators and quantifiers of the axiom, the top-level phenotype can be represented as a nested AND/OR expression including the elements that are supported at a specific site. However, before generating a CDS-level query, the Phenotype Converter always checks if a phenotype-level mapping is available for all or a part of a phenotype tree. If yes, the high-level mapping is used in preference to the CDS-level mapping. This means that the name of the phenotype will appear in the SQL query instead of its CDS-level representation. As an example, consider the fragment of the NEUROWEB Ontology in Fig. 7; it represents a simplified version of Presence of Carotid Stenosis concept as defined by the NEUROWEB clinicians. In a patient the presence of a stenosis in the carotid artery may be diagnosed, if there is evidence of it in the segment called Internal/ Common Carotid Artery (ICA–CCA) or in the Origin of the Common Carotid Artery. The ICA–CCA Stenosis can be diagnosed if there is evidence of it in the right or in the left Internal/Common Carotid Artery and in particular if there is an evidence of a severe stenosis or a complete occlusion of the vessel. One of the two branches reaching the last level of the ontology states that the occlusion must be diagnosed by means of the CDS indicator Duplex Carotid Degree Of Stenosis in Left CCA–ICA, and that this indicator must have value (Occlusion). For every clinical site, the mapping can be done at the best fitting level: for example if the clinical database ‘‘A” includes information only about the presence of a carotid stenosis in the patient, but without details on its position, the mapping can be done only at the phenotype level, but not at the CDS level. Thanks to the phenotype-level mapping, a generic query that asks for patients with carotid stenosis will receive reliable answers. For more specific queries, for example select all patients with a stenosis in the Internal Carotid Artery, the database ‘‘A” cannot provide answers due to a lack of information.
Fig. 7. A visual diagram for the Presence of Carotid Occlusion as represented in the Reference Ontology. Big boxes represent entities, arrows represent relations. Small gray boxes (transitions) are used to convey AND/OR logical connections: AND corresponds to relations using the same transition, OR corresponds to relation using independent transitions.
G. Colombo et al. / Journal of Biomedical Informatics 43 (2010) 469–484
479
Fig. 8. The figure displays the information flow of a clinical query in the NEUROWEB system. The software modules are displayed as solid boxes, whereas the data items are displayed as cylinders. The dashed boxes group the components of a local site (only two sites were depicted for reasons of compactness). Thick arrows identify the information flow elicited by each user’s query.
The system has to build a targeted query for each database, depending on the level of detail at which it is mapped. The Phenotype Converter performs this task. In the previous example, a query select all patients with presence of carotid stenosis = Yes will easily receive answers from the database ‘‘A”, but it must also receive answers from a database ‘‘B” that supports mapping at the CDS level. The Phenotype Converter translates the previous query into select all patients with Duplex Carotid Degree Of Stenosis in Left CCA–ICA = Occlusion OR Duplex Carotid Degree Of Stenosis in Left CCA–ICA = Severe OR Duplex Carotid Degree Of Stenosis in Right CCA–ICA = Severe OR . . . OR Duplex Carotid Degree Of Stenosis in Origin Common = Severe’, by following the Ontology axioms that state how each concept is defined in terms of CDS elements. Of course, a database ‘‘C” can be mapped at any intermediate level. For example, if the distinction between the ICA–CCA and the Origin Common Carotid Artery were present in database ‘‘C”, the mapping would occur at that level, and the query would be translated in select all patients with ICA–CCA Stenosis = Yes OR Origin Common Carotid Artery Stenosis = Yes. The NEUROWEB system exploits well-established Web services technology to decouple the central components from the local sites in order to facilitate access by new clinical partners. Although each partner is free to chose the preferred technology to implement the local components, we have developed and made available a reference implementation to deliver a sample solution that exploits the popular open-source technologies Glassfish for the communication tasks [70] and Postgres [71] for implementing the database view. Glassfish safely manages call wrappings to expose the local interface as a WSDL document. The Clinical Query Application runs as a web application on the Glassfish server, processes the incoming web service call, translates it into a SQL query, runs the query on the database view, wraps and returns the resulting record set.
4. Conclusions The purpose of the NEUROWEB Reference Ontology is to support the retrieval of patient clusters in a rich semantic context, coupled with strict requirements on the clarity and coherence of clustering criteria. In fact, the criteria adopted to categorize patients in computer-unaided practice are deeply rooted in diagnostic knowledge of the cerebrovascular experts. In addition, the specific commitment to association studies is particularly demanding from a methodological standpoint: on one hand, it is effective to provide the different researchers with a set of common and explicit clustering criteria, endued with a shared semantic; on the other hand, it is necessary to grant coherence and consistency when different repositories are integrated into a unified, virtual data warehouse. The tremendous importance of such issues is the rationale driving the cerebrovascular community to establish standards for stroke classification, such as the TOAST [39–41]; in particular, the growing amount of available data requires the adoption of computer-aided practices, and thus formal representation of classification systems. The specific advantages provided by an ontological modeling approach with respect to the issues discussed are clear. On one hand, the explicit representation of classification criteria, as provided by the layer of the Low Phenotypes, enables to analytically express the meaning of the groupings organized in the Top Phenotype layer. On the other hand, the mapping to the local clinical repositories, mediated by the Core Data Set layer, enables an automated retrieval of patient clusters driven by the previously defined criteria, which spares the researcher from tedious and time-consuming query formulation activities, though preserving methodological coherence. Considering this latter issue, the main problem is that different repositories contain classification data, but it is not granted that they were assigned applying coherent criteria.
480
G. Colombo et al. / Journal of Biomedical Informatics 43 (2010) 469–484
Moving to a more general level, the NEUROWEB ontological framework provides a formalization of concepts explicitly or implicitly encoded in the TOAST system, a scope not addressed by the available medical ontologies. According to this perspective, the NEUROWEB modeling effort, in spite of its prototypical nature, covers an unchallenged area of medical knowledge representation, which can be reasonably deemed to elicit the interest of the broader cerebrovascular community. An even greater challenge is posed by the analysis and validation of genotype–phenotype associations, which will be generated by high-throughput genotyping campaigns on NEUROWEB patients. Nevertheless, the present structure adopted for the ontological model proved to be already adequate for the extension to the biomolecular domain, providing an effective groundwork to establish relations between cerebrovascular pathologies and their underlying molecular mechanisms. As a whole, the Low Phenotype layer, by analytically deconstructing the stroke types into general concepts of the cerebrovascular domain, provides an effective groundwork both for (a) the phenotype redefinition, and (b) the representation of relations between
cerebrovascular pathologies and their underlying molecular mechanisms.
Acknowledgments The authors wish to thank all the clinical partners: Istituto Nazionale Neurologico Carlo Besta (INNCB, Milan – Italy), Orszagos Pszichiatriai es Neurologiai Intezet (AOK-OPNI, Budapest – Hungary), University of Patras (UOP, Patras – Greece), Erasmus Universitair Medisch Centrum Rotterdam (MI-EMC, Rotterdam – Holland). In particular, we wish to thank Dr. Stella Marousi from UOP, Dr. Csaba Ovary from AOK-OPNI, Dr. Philip Homburg from MI-EMC, and Dr. Eugenio Parati from INNCB, for their valuable contributions during the knowledge acquisition campaign and model refinement process. Finally, the authors wish to express the utmost gratitude to Prof. Giancarlo Mauri from the Dipartimento di Informatica Sistemistica e Comunicazione (DISCo) of the Università degli Studi di Milano Bicocca for his coordination efforts and leadership role in the NEUROWEB project.
Fig. A.1. Specification of Data sources.
Fig. A.2. The Phenotype Specification GUI.
G. Colombo et al. / Journal of Biomedical Informatics 43 (2010) 469–484 Table A.1 Example A. The query compiled for Clinic A (with full CDS-level mapping). SELECT //patient ID, selected automatically ‘‘A.01.01.01.00.00.00”, //the CDS codes for the 5 elements specified in step 5 ‘‘A.01.01.02.00.00.00”, ‘‘A.01.02.01.01.00.00”, ‘‘A.01.02.09.02.01.00”, ‘‘D.01.01.01.00.00.00”, ‘‘A.01.01.07.00.00.00” FROM
WHERE ‘‘A.01.01.01.03.00.00” = 1 //meaning gender = female AND ( //definition of Atherosclerotic_Ischemic_Stroke ( //definition of Ischemic_Stroke, //involving 11 CDS-level constraints, NOT DETAILED HERE ... ) AND ( //definition of Atherosclerotic_Disease, //involving 5 CDS-level constraints ‘‘A.01.02.02.10.01.00” = 2 //myocardial infarction ‘‘OR A.01.02.02.11.01.00” = 2 //angina pectoris ‘‘OR A.01.02.02.09.01.00” = 2 //family stroke/TIA ‘‘OR A.01.02.02.05.01.00” = 2 //arterial hypertension ‘‘OR A.01.02.02.07.01.00” = 2 //hypercholesterolemia ) )
Table A.2 Example B. The query compiled for Clinic B (with a phenotype-level mapping). SELECT //patient ID, selected automatically ‘‘A.01.01.01.00.00.00”, //the CDS codes for the 5 elements specified in step 5 ‘‘‘‘A.01.01.02.00.00.00”, ‘‘‘‘A.01.02.01.01.00.00”, ‘‘‘‘A.01.02.09.02.01.00”, ‘‘‘‘D.01.01.01.00.00.00”, ‘‘‘‘A.01.01.07.00.00.00” FROM WHERE ‘‘A.01.01.01.03.00.00” = 1 //meaning gender = female AND ( //definition of Atherosclerotic_Ischemic_Stroke ( //definition of Ischemic_Stroke, //involving 11 CDS-level constraints, NOT DETAILED HERE ... ) AND ( //definition of Atherosclerotic_Disease, //involving a single phenotype-level mapping Atherosclerotic_Disease = ‘yes’ ) )
Appendix A A.1. Exploiting the Reference Ontology in a concrete test case This section exemplifies the use of the NEUROWEB Reference Ontology for queries as well as the interactions between the user and the system interface. We will focus on the top phenotype Atherosclerotic_Ischemic_Stroke, which is defined as the intersection of Ischemic_Stroke and Atherosclerotic_Disease. Specifically, the presence of Atherosclerotic_Disease in a patient is defined in the Reference Ontology as a disjunction of five CDS-level constraints (CDS alphanumeric codes are displayed between parentheses). Presence of myocardial infarction = yes, within 4 weeks (A.01.02.02.10.01.00 = 2). Presence of angina pectoris = yes (A.01.02.02.11.01.00 = 2).
481
Presence of family stroke/TIA = yes (A.01.02.02.09.01.00 = 2) Presence of arterial hypertension = yes (A.01.02.02.05.01.00 = 2). Presence of hypercholesterolemia = yes (A.01.02.02.07.01.00 = 2). The structure of this phenotype is shown in the lower right part of Fig. A.2. We omit the definition of Ischemic Stroke only for simplicity. Now we show a typical use of the NEUROWEB system, focusing on how the reference ontology is exploited for clinical, genomic, and literature searches. (1) Define the mapping. Suppose that two NEUROWEB clinics (clinic A and clinic B) compile the NEUROWEB-to-Local mapping for a subset of the CDS elements. Clinic A has full CDS mapping, whereas clinic B can support mapping only at the phenotype level for Atherosclerotic disease. The list of supported CDS elements and phenotypes is sent and stored in the central database. The clinics also set up their local web services. (2) Define the goal. A NEUROWEB user, with read access granted to the databases of clinic A and B, intends to study how atherosclerotic ischemic stroke in women might be associated with genetic polymorphisms. (3) Select a data source. The user enters the NEUROWEB portal and selects clinic A and clinic B as data sources. If we consider the whole query as SELECT data elements of interest FROM clinical DBs of interest WHERE patient filter condition, then this step specifies the FROM part (Fig. A.1). (4) Specify a phenotype. The user moves to the next tab of the GUI to specify a phenotype. She selects the phenotype Atherosclerotic ischemic stroke. The tree structure, with the CDS elements as leaf nodes, appears in the right pane (see Fig. A.2) The tree shows the logical relations (AND/OR) of its branches as well as the required data values at the CDS elements. Note that the top phenotype Atherosclerotic_Ischemic_Stroke is defined as Ischemic_Stroke AND Atherosclerotic_Disease. Fig. A.2 does not show the structure of the low phenotype Ischemic_Stroke.The user can browse and modify the phenotype tree at any level, for example, she can change the value of CDS element Presence of angina pectoris from ‘‘yes” to ‘‘no”. The GUI supports any combination of existing phenotypes and CDS elements to form a new phenotype, which the user can save as her own phenotype for future use. Now she wants to specify female patients, so she adds the CDS element ‘‘Gender” and sets its value to ‘‘Female”, then connects it to the phenotype tree with an AND relation. This action is shown in Fig. A.2.The GUI reads the Reference Ontology to display the tree. The user-compiled phenotype will later be the source of the WHERE part of the query. (5) Select the dataset elements. The user selects on the next tab the CDS elements or phenotypes that she wants to view for the patients matching the user-defined phenotype. This list will be the SELECT part of the query. If this list contains an item that is unsupported for a certain clinic, the returned values will be null. Suppose that the user selects: Date of birth (CDS code A.01.01.02.00.00.00). Date of admission (CDS code A.01.02.01.01.00.00). Discharge status (CDS code A.01.02.09.02.01.00). Time between onset and first scan (CDS code D.01.01.01.00.00.00). Marital status (CDS code A.01.01.07.00.00.00). (6) Running the query. The user can set some additional parameters and then execute the query. The phenotype is sent to the Phenotype Converter that, using the mapping informa-
482
G. Colombo et al. / Journal of Biomedical Informatics 43 (2010) 469–484
tion, produces two different WHERE parts for the clinics A and B. For clinic A, the query will contain the AND/OR combination of the 16 CDS-level constraints that make up the phenotype, plus the constraint on gender. See Table A.1. For clinic B, the query will contain Atherosclerotic disease = ‘yes’ instead of the 5 CDS-level constraints that define Atherosclerotic disease in the Reference Ontology. This clause is combined with several other CDS constraints like Presence of angina pectoris = ‘no’ etc., coming from the other parts of the user-defined phenotype tree. See Table A.2 The two queries, with the same SELECT parts, are sent to the clinical web services declared in the FROM part, and the matching records are returned (Fig. A.3).
(7) Analyze the result set. The user can browse aggregate and detail information on the patients in tabular form. She can run data mining methods like decision tree to select the dominant variables of the SELECT part, with respect to any variable chosen as class label. She can also save the complete query as well as the returned patient set to ensure consistency with future queries in longitudinal studies. (8) Run a semantic query. The goal here is to find publications that link the user-defined phenotype to a chromosome. When started, the Semantic Engine searches a cached copy of public resources like MedLine to find publications relevant to the phenotype specified by the user. The engine extracts search terms from the phenotype tree, also allowing the user to edit them. Then it runs the search, using the Reference Ontology also to rewrite the term list in case of too
Fig. A.3. The patient set returned as a response, alongside aggregated data (upper part) and individual patient records (lower part).
Fig. A.4. Semantic Query definition GUI.
G. Colombo et al. / Journal of Biomedical Informatics 43 (2010) 469–484
few or too many hits. The user views the ranked and highlighted abstracts, and may refine the original phenotype accordingly (see Fig. A.4). (9) Run a genomic query. Suppose the locus 9p21 was suggested by the relevant publications in the previous step. Now, using the Genomic Engine, the user can search public genomic databases to find out which SNPs are associated to this locus, and then check which of these her patients in the returned record set actually have, thus verifying the original hypothesis. This concludes the phenotype–genotype association study. The above example demonstrated a phenotype–genotype study. The NEUROWEB tools, however, can be used for other purposes like verifying a suggested treatment in the partners’ databases for a certain phenotype or a single patient.
[23]
[24]
[25]
[26] [27]
[28] [29] [30]
References [1] Botstein D, Risch NJ. Discovering genotypes underlying human phenotypes: past successes for mendelian disease, future approaches for complex disease. Nat Genet 2003;33(Suppl.):228–37. [2] Tabor HK, Risch NJ, Myers RM. Candidate-gene approaches for studying genetic traits: practical considerations. Nat Rev Genet 2002;3(5):391–7. [3] Beneventano D, Bergamaschi S, Guerra F, Vincini M. Synthesizing an integrated ontology. IEEE Internet Comput 2003;7(5):42–51. [4] Beneventano D, Bergamaschi S, Lodi S, Sartori C. Consistency checking in complex object database schemata with integrity constraints. IEEE Trans Knowledge Data Eng 1998;10(4):576–98. [5] Lenz R, Beyer M, Kuhn KA. Semantic integration in healthcare networks. Int J Med Inform 2006;76(2–3):201–7. [6] Brazhnik O, Jones JF. Anatomy of data integration. J Biomed Inform 2007;40(3):252–69. [7] Rahm E, Bernstein PA. A survey of approaches to automatica schema matching. VLDB J 2001;10(4):334–50. [8] Pérez-Rey D, Maojo V, Garcia-Remesal M, Alonso-Calvo R, Billhardt H, MartinSánchez F, et al. ONTOFUSION: ontology-based integration of genomic and clinical databases. Comput Biol Med 2006;36(7):712–30. [9] Sun Y. Methods for automated concept mapping between medical databases. J Biomed Inform 2004;37(3):162–78. [10] Alladin-eHealth Project. Document at <www.alladin-ehealth.org/Pub/Docs/ D9.1.pdf>; 2007. [11] Arens Y, Hsu C-N, Knoblock CA. Query processing in the SIMS information mediator. In: Readings in agents. Morgan Kaufmann Publishers Inc.; 1998. p. 82–90. [12] Sun J, Miao H, Cao X. A domain formal ontology and the application in service component retrieval. In: Proceedings of the international conference on software engineering advances (ICSEA), Tahiti, French Polynesia. IEEE Computer Society; 2006. [13] Garcia-Remesal M, Maojo V, Billhardt H, Crespo J, Alonso-Calvo R, Perez-Rey D, et al. ARMEDA II: supporting genomic medicine through the integration of medical and genetic databases. In: Proceedings of the fourth IEEE symposium on bioinformatics and bioengineering (BIBE’04). IEEE Computer Society; 2004. [14] Köhler J, Philippi S, Lange M. Semeda: ontology based semantic integration of biological databases. Bioinformatics 2003;19(18):2420–7. [15] Mougin F, Burgun A, Bodenreider O, Chabalier J, Loréal O, Beux P. Automatic methods for integrating biomedical data sources in a mediator-based system. In: DILS ’08: Proceedings of the fifth international workshop on data integration in the life sciences. Heidelberg, Berlin: Springer-Verlag; 2008. p. 61–76. [16] Sahoo SS, Bodenreider O, Rutter JL, Skinner KJ, Sheth AP. An ontology-driven semantic mashup of gene and biological pathway information: application to the domain of nicotine dependence. J Biomed Inform 2008;41(5):752–65. [17] Pérez-Rey D, Maojo V, García-Remesal M, Alonso-Calvo R, Billhardt H, MartinSánchez F, et al. ONTOFUSION: ontology-based integration of genomic and clinical databases. Comput Biol Med 2005. [18] Coulet A, Smaı¨l-Tabbone M, Benlian P, Napoli A, Devignes M-D. Ontologyguided data preparation for discovering genotype–phenotype relationships. BMC Bioinform 2008;9(S-4). [19] Rubin DL, Lewis SE, Mungall CJ, Misra S, Westerfield M, Ashburner M, et al. National Center for Biomedical Ontology: advancing biomedicine through structured organization of scientific knowledge. OMICS 2006;10(2):185–98. [20] Coulet A, Smaı¨l-Tabbone M, Benlian P, Napoli A, Devignes M-D. SNP-converter: an ontology-based solution to reconcile heterogeneous SNP descriptions for pharmacogenomic studies. In: DILS; 2006. p. 82–93. [21] Burgun A, Bodenreider O. Accessing and integrating data and knowledge for biomedical research. Yearbook Med Inform 2008:91–101. [22] So H-C, Chen RYL, Chen EYH, Cheung EFC, Li T, Sham PC. An association study of rgs4 polymorphisms with clinical phenotypes of schizophrenia in a chinese
[31] [32]
[33]
[34] [35]
[36]
[37] [38] [39]
[40]
[41]
[42]
[43] [44] [45]
[46]
[47] [48]
[49] [50] [51] [52] [53] [54]
483
population. Am J Med Genet Part B, Neuropsychiatr Genet 2008;147B(1):77–85. Lee C, Kong M. An interactive association of common sequence variants in the neuropeptide y gene with susceptibility to ischemic stroke. Stroke 2007;38(10):2663–9. Boncoraglio GB, Bodini A, Brambilla C, Carriero MR, Ciusani E, Parati EA. An effect of the pai-1 4g/5g polymorphism on cholesterol levels may explain conflicting associations with myocardial infarction and stroke. Cerebrovasc Dis 2006;22(2–3):191–5. Scheuermann RH, Ceuster W, Smith B. Towards an ontological treatment of disease and diagnosis, cf. ; March 2009, in press. Guarino N, Poli R, editors. Formal ontology in conceptual analysis and knowledge representation. Kluwer Academic Press; 1994. Bouquet P, Donà A, Serafini L, Zanobini S. ConTeXtualized local ontology specification via CTXML. In: MeaN-02 – AAAI workshop on meaning negotiation, Edmonton, Alberta, Canada. AAAI; 2002. Van der Pet PE, Mars NJI. Bottom-up construction of ontologies. IEEE Trans Knowledge Data Eng 1998;10(4):513–26. Bard JB, Rhee L, Seung Y. Ontology in biology: design, application and future challenges. Nat Rev Genet 2004;3:213–22. Fielding JM, Simon J, Ceusters W, Smith B. Ontological theory for ontological engineering: biomedical systems information integration. In: Proceedings of the ninth international conference on the principles of knowledge representation and reasoning (KR2004), Whistler, BC, Canada; June 2004. Gomez-Perez A, Corcho O, Fernandez-Lopez M. Ontological engineering. New York, NY, USA: Springer-Verlag; 2004. Scheuermann R, Smith B. Signs, symptoms and findings: first steps toward an ontology of clinical phenotypes. Web site at <www.bioontology.org/wiki/ index.php/DallasWorkshop>. Dallas Phenotype Workshop; September 2008. Smith C, Goldsmith CA, Eppig J. The mammalian phenotype ontology as a tool for annotating, analyzing and comparing phenotypic information. Genome Biol 2004;6(1):R7. PATO – Phenotypic Quality Ontology. Web site at <www.bioontology.org/wiki/ index.php/PATO:MainPage>. Bodenreider O, Smith B, Kumar A, Burgun A. Investigating subsumption in SNOMED-CT: an exploration into large description logic-based biomedical terminologies. Artif Intell Med 2007;39(3):183–95. Linksvan der Kooij J, Goossen WT, Goossen-Baremans AT, de Jong-Fintelman M, van Beek L. Using SNOMED-CT codes for coding information in electronic health records for stroke patients. Stud Health Technol Inform 2006;124:815–23. Cimino JJ. Coding systems in health care. Methods Inform Med 1996;35(4– 5):273–84 [Review paper]. Disease Ontology (DisO) – The NUgene Project. Web site at . Adams Jr HP, Bendixen BH, Kappelle JJ, Biller J, Love BB, Gordon DL, et al. Classification of subtype of acute ischemic stroke, definition for use in a multicenter clinical trial, TOAST. Trial of Org 10172 in Acute Stroke Treatment. Stroke 1993;24:35–41. Goldstein LB, Jones MR, Matchar DB, Edwards LJ, Hoff J, Chilukuri V, et al. Improving the reliability of stroke subgroup classification using the Trial of ORG 10172 in Acute Stroke Treatment (toast) criteria. Stroke 2001;32: 1091–7. Ay H, Furie KL, Singhal A, Smith WS, Sorensen AG, Koroshetz WJ. An evidencebased causative classification system for Acute Ischemic Stroke. Ann Neurol 2005;58:688–97. Taylor CF, Field D, Sansone S-A, Aerts J, Apweiler R, Ashburner M, et al. Promoting coherent minimum reporting guidelines for biological and biomedical investigations: the MIBBI project. Nat Biotechnol 2008;26(8):889–96. Donnelly K. SNOMED-CT: the advanced terminology and coding system for eHealth. Stud Health Technol Inform 2006;121:279–90. Lipscomb CE. Medical Subject Headings (MeSH). Bull Med Library Assoc 2000;88(3). Stenson PD, Ball EV, Mort M, Phillips AD, Shiel JA, Thomas NS, et al. Human Gene Mutation Database (HGMD): 2003 update. Human Mutat 2003;21(6):577–81. Stenson PD, Ball E, Howells K, Phillips A, Mort M, Cooper DN. Human Gene Mutation Database: towards a comprehensive central mutation database. J Med Genet 2008;45(2):24–126. Lewis SE. Gene Ontology: looking backwards and forward. Genome Biol 2004;6(103). Spasic I, Ananiadu S, McNaught J, Kumar A. Text mining and ontologies in biomedicine: making sense of raw text. Brief Bioinform 2005;6(3): 239–51. Chen Y, Perl Y, Geller J, Cimino JJ. Analysis of a study of the users, uses, and future agenda of the UMLS. J Maerican Med Inform Assoc 2007;14(2):221–31. ICD9-CM. Web site at <www.icd9cm.net>. ICD10. Web site at <www.who.int/classifications/icd>. Fellbaum C, editor. WordNet: an electronic lexical database. MIT Press; 1998. Jensen CS, Snograss RT. Temporal data management. Knowledge Data Eng 1999;11(1):36–44. Artale A, Parent C, Spaccapietra S. Modeling the evolution of objects in temporal information systems. In: Kirchberg M, Hegner SJ, editors, Proceedings of the fourth international symposium of foundations of
484
[55]
[56]
[57]
[58] [59] [60] [61]
G. Colombo et al. / Journal of Biomedical Informatics 43 (2010) 469–484 information and knowledge-based systems. Lecture notes in computer science, vol. 3861; 2006. p. 22–42. Shahar Y. Timing is everything: temporal reasoning and temporal data maintenance in medicine. In: Joint European conference on artificial intelligence in medicine and medical decision making (AIMDM’99). Lecture notes in artificial intelligence, vol. 1620; 1999. p. 30–46. Rector AL, Rogers J. Ontological and practical issues in using a description logic to represent medical concept systems: experiences from GALEN. In: Reasoning Web 2006. Lecture notes in computer science, vol. 4126. Springer-Verlag; 2006. p. 197–231. Palma P, Llamas B, Gonzales A, Menarguez M. Acquisition and representation of causal and temporal knowledge in medical domains. In: Knowledge-based intelligent information and engineering systems, seventh international conference, KES 2003, Lecture notes in computer science, vol. 2777; 2003. p. 1284–90. International HapMap Consortium. A haplotype map of the human genome. Nature 2005;437(7063):1299–320. Sherry ST, Ward MH, Kholodov M, Baker J, Phan L, Smigielski EM, et al. dbSNP: the NCBI database of genetic variation. Nucleic Acids Res 2001;29:308–11. Meschia JF. Clinically translated ischemic stroke genomics. Stroke 2004;35(11 Suppl. 1):2735–9. Kaneisha M, Goto S. KEGG: Kyoto encyclopedia of genes and genomes. Nucleic Acid Res 2000;28:27–30.
[62] Kaneisha M, Goto S, Hattori M, Aoki-Kinoshita KF, Itoh M, Kawashima M, et al. From genomics to chemical genomics: new developments in KEGG. Nucleic Acid Res 2006;34:D354–7. [63] Kaneisha M, Araki M, Goto S, Hattori M, Hirakawa M, Itoh M, et al. KEGG for linking genomes to life and the environment. Nucleic Acid Res 2008;36:D480–4. [64] Becker K, Barnes KC, Bright TJ, Wang SA. The genetic association database. Nat Genet 2004;36(5):431–2. [65] Baader F, Calvanese D, McGuinness DL, Nardi D, Patel-Schneider PF, editors. The description logic handbook: theory, implementation and application. New York, NY, USA: Cambridge University Press; 2003. [66] Sattler U. Description logics for the representation of aggregate objects. In: Proceedings of the 14th European conference on artificial intelligence. Amsterdam, Holland: IOS Press; 2000. [67] Horrocks I, Sattler U, Tobies S. Practical reasoning for expressive description logics. In: Proceedings of the sixth international conference on logic and automated programming (LPAR99), Tblisi, Georgia, 1999. [68] W3C. OWL Web Ontology Language. Web site at <www.w3c.org/TR/owlguide>, February 2004. [69] Jena – A Semantic Web Framework for Java. Web site at . [70] GlassFish. Web site at . [71] Douglas K, Douglas S. PostgreSQL. Sams, 2nd ed.; 2005.
This article appeared in a journal published by Elsevier. The attached copy is furnished to the author for internal non-commercial research and education use, including for instruction at the authors institution and sharing with colleagues. Other uses, including reproduction and distribution, or selling or licensing copies, or posting to personal, institutional or third party websites are prohibited. In most cases authors are permitted to post their version of the article (e.g. in Word or Tex form) to their personal website or institutional repository. Authors requiring further information regarding Elsevier’s archiving and manuscript policies are encouraged to visit: http://www.elsevier.com/copyright
Author's personal copy Letters to the Editor
137
Fig. 1. Comparison of factors influencing the time from arrival to the occurrence of AMI: a. Male sex (p = 0.01), b. Low training status (p b 0.05), c. Altitude N1200 m (p = 0.01). [3] Bhaskaran K, Hajat S, Haines A, Herrett E, Wilkinson P, Smeeth L. Effects of ambient temperature on the incidence of myocardial infarction. Heart 2009;95(21):1760–9. [4] Buskirk ER, Kollias J, Akers RF, Prokop EK, Reategui EP. Maximal performance at altitude and on return from altitude in conditioned runners. J Appl Physiol 1967;23(2):259–66. [5] Tofler GH, Muller JE. Triggering of acute cardiovascular disease and potential preventive strategies. Circulation 2006;114(17):1863–72. [6] Mittleman MA, Maclure M, Sherwood JB, Mulry RP, Tofler GH, Jacobs SC, et al. Triggering of acute myocardial infarction onset by episodes of anger. Determinants of Myocardial Infarction Onset Study Investigators. Circulation 1995;92(7):1720–5. [7] D'Agostino Sr RB, Vasan RS, Pencina MJ, Wolf PA, Cobain M, Massaro JM, et al. General cardiovascular risk profile for use in primary care: the Framingham Heart Study. Circulation 2008;117(6):743–53. [8] Graham I, Atar D, Borch-Johnsen K, Boysen G, Burell G, Cifkova R, et al. European guidelines on cardiovascular disease prevention in clinical practice: full text. Fourth Joint Task Force of the European Society of Cardiology and other societies on cardiovascular disease prevention in clinical practice (constituted by representatives of nine societies and by invited experts). Eur J Cardiovasc Prev Rehabil 2007;14(Suppl 2):S1–S113.
[9] Mittleman MA, Maclure M, Tofler GH, Sherwood JB, Goldberg RJ, Muller JE. Triggering of acute myocardial infarction by heavy physical exertion. Protection against triggering by regular exertion. Determinants of Myocardial Infarction Onset Study Investigators. N Engl J Med 1993;329(23):1677–83. [10] Schmid JP, Noveanu M, Gaillet R, Hellige G, Wahl A, Saner H. Safety and exercise tolerance of acute high altitude exposure (3454 m) among patients with coronary artery disease. Heart 2006;92(7):921–5. [11] Klug G, Zwick RH, Frick M, Wolf C, Schocke MF, Conci E, et al. Impact of exercise capacity on myocardial high-energy phosphate metabolism. Int J Sports Med 2007;28(8):667–72. [12] Burtscher M, Faulhaber M, Flatz M, Likar R, Nachbauer W. Effects of short-term acclimatization to altitude (3200 m) on aerobic and anaerobic exercise performance. Int J Sports Med 2006;27(8):629–35. [13] Cold exposure and winter mortality from ischaemic heart disease, cerebrovascular disease, respiratory disease, and all causes in warm and cold regions of Europe. The Eurowinter Group. Lancet 1997;349(9062):1341–6. [14] Shewan L, Coats A. Ethics in the authorship and publishing of scientific articles. Int J Cardiol 2010;144:1–2.
0167-5273/$ – see front matter © 2011 Elsevier Ireland Ltd. All rights reserved. doi:10.1016/j.ijcard.2011.02.005
Stress ECG utilization in the evaluation of patients with chest pain: The real practice in Hungary with 10 million inhabitants István Kósa a,⁎, István Vassányi a, Attila Nemes b, József Hortobágyi c, György Kozmann a a b c
Research and Developing Center for Health Informatics, University Pannonia, Veszprém, Hungary 2nd Department of Medicine and Cardiology Center, Medical Faculty, Albert Szent-Györgyi Clinical Center, University of Szeged, Szeged, Hungary National Institute of Strategic Health Research, Budapest, Hungary
a r t i c l e
i n f o
Article history: Received 2 February 2011 Accepted 6 February 2011 Available online 26 February 2011 Keywords: Stress ECG Evaluation Chest pain Practice
⁎ Corresponding author at: Research and Developing Center for Health Informatics, University Pannonia, H-8200, Egyetem street 10, Veszprém, Hungary. Tel.: +36 70 3201192; fax: + 36 88 624526. E-mail address: [email protected] (I. Kósa).
The latest NICE guidelines for the diagnosis of chest pain inspired hot debates regarding the usefulness of stress ECG in the evaluation of these patients [1–4]. To bridge the gap between this statement and the current practice, Nijjer et al. performed a survey in the UK, which was published recently in the Journal [5]. In order to give another insight in the daily routine, we analysed the typical patient pathways based on the reimbursement databases in Hungary, a country that has a single health care funding organization. Our purpose was to survey the weight of individual diagnostic tools in this process and characterise the subpopulations following the individual pathways, based on the survival data. We included 639,139 patients in our analysis from the database of National Institute of Strategic Health Research who underwent Ischemic Heart Disease (IHD) related diagnostic procedures between 1 January 2004 and 31 December 2008 in an outpatient or inpatient care. The death registry follow-up data of these patients was available until 31 December 2009. We labelled each inpatient event based on
Author's personal copy 138
Letters to the Editor
Table 1 Inpatient events, related International Classification of Diseases (ICD) codes, and labels used in the analysis.
Table 3 Outpatient events, related International Classification of Procedures in Medicine (ICPM) codes, and labels used in the analysis.
Case type
ICD codes
Case type
ICPM codes
Label
Stabile ischemic heart disease Angina Acute myocardial infarction hospitalization
I2480, I2500, I2510, I2520, I2530, I2540, I2550, ‘b’ I2560, I2580, I2590 I2000, I2010, I2080, I2090, I2490 ‘a’ I2100, I2110, I2120, I2130, I2140, I2190, I2191, I2200, ‘i’ I2210, I2280, I2290, I2300, I2310, I2320, I2330, I2340, I2350, I2360, I2380, I2400, I2410
Stress ECG Stress myocardial perfusion SPECT Stress Echo
12620, 12621, 12653 3521D, 35221, 35223, 3562A, 3581A, 3581B
‘e’ ‘s’
3612M
‘o’
Label
the International Classification of Diseases (ICD) codes as shown in Table 1. If a case has had diagnosis codes belonging to two or more of the groups, the case was classified using a priority order of ‘i–a–b’. Inpatient cases were reclassified if they had a relevant International Classification of Procedures in Medicine (ICPM) code, distinguishing cases with acute myocardial infarction (AMI) hospitalization. Therefore, the reclassification process resulted in different labelling for cases with and without AMI as presented in Table 2. Similarly, outpatient cases were classified based on the ICPM codes as shown in Table 3. Based on the death registry we created an ‘h’ labelled event from each death case. As a result, our ‘raw classified event list’ consisted of 1,992,942 events. Since the clinical coding practice involves some interventions with multiple ICPM codes, generating multiple events (e.g. ‘e’ + ‘s’ for stress SPECT), we merged the same-day, same-place events based on a priority order ‘h–j–c–q–p–r–n–g–k–i–o–s–e–a–b’ into a single event in a recursive manner. To eliminate the event duplicates for the two-day SPECT protocol, a similar second merging cycle was performed for event sequences ‘s–s’, if the events happened within 7 days. The two merging steps reduced the event count to 1,295,452. In order to analyse patterns in the event sequences, we formed event strings for each patient by concatenating the event labels and tags representing follow-up times (in days) between two consequent events, into a single string. To focus our analysis on patients with stable conditions we considered only patients with at least 180 days of event-free period before the first labelled event. This event was qualified as an index event if: • it was an outpatient diagnostic test with label ‘e’,‘s’ or , ‘o’ or • it was an IHD hospitalization followed by invasive procedure (event ‘a’ or ‘b’ followed by ‘k’ or ‘p’ within 180 days), or • it was an invasive procedure excluding those preceding nonrevascularising invasive procedures (events ‘k’ or ‘g’ not followed by ‘n’ or ‘r’ within 180 days). If a suitable index event was found, the subsequent events within 365 days were analysed. For each of such event we computed the ‘event follow-up time’ as the time between the index event and the ending date of the follow-up period (31 December 2009), or, if the patient died, the date of the death. Out of all patients, 92,495 had no suitable index event or had an invalid date of birth, and were thus excluded from further analysis. The remaining 546,644 patients had a total of 818,568 events during the follow-up windows. We created patient groups based on the index
event as well as on the first two events. The initial (index) event was stress ECG (‘e’) in 433,032 cases (79.22%), stress SPECT (‘s’) in 25,976 cases (4.75%), stress Echo (‘o’) in 3377 cases (0.62%), elective coronary angiography (‘k’ or ‘p’) in 29,172 cases (5.34%), coronary angiography in AMI patients (‘g’ or ‘q’) in 17,690 cases (3.24%), and IHD hospitalization followed by an invasive assessment (‘a’ or ‘b’) in 37,397 cases (6.84%). If stress ECG was the initial event, there was no further investigation during the one-year follow-up in 380,083 cases (87.77% of the branch), while it was followed by elective coronary angiography in 14,977 cases (3.46% of the branch). One-year survival curves of the individual branches are presented in Fig. 1. Our data shows that stress ECG dominates so heavily the evaluation of chest pain patients, that it is hard to believe that it could be replaced by imaging techniques in the near future. Stress ECG does have a discriminative capability for very low versus low risk patients, although patients referred to invasive procedure after a stress ECG still have lower mortality risk than subjects for stress imaging as initial investigation. The method used for this evaluation has limitations, but also has some indubitable strength. The data set is almost complete for the population. Outpatient care procedures are depicted with high reliability, because ICPM-based financing does not allow any overand under reporting. The current ICPM code list does not contain new procedures, performed only in some institutes, like CT coronary angiography or cardiac MR. However, due to the limited number of these investigations in our country at the time of data sampling, this fact hardly ever influences our conclusions. The analysed invasive procedures in the inpatient care also have financing consequences in the applied Diagnosis-Related Group financing system, which warrant their reliability. The noninvasive procedures without such financial consequences in the inpatient care were excluded from our analysis to avoid biased sampling. Since patients with stable chest pain are evaluated dominantly on an outpatient basis and only patients with unstable condition are referred for hospital assessment, this restriction is considered to be acceptable. IHD ICD ICPM AMI
Ischemic Heart Disease International Classification of Diseases International Classification of Procedures in Medicine Acute Myocardial Infarction
We gratefully acknowledge the skilled assistance of the staff of the National Institute of Strategic Health Research (Head: György Surján, MD, PhD). The authors of this manuscript have certified that they comply with the Principles of Ethical Publishing in the International Journal of Cardiology [6].
Table 2 Inpatient events, related International Classification of Procedures in Medicine (ICPM) codes, and labels used in the analysis. Case type
ICPM codes
Label w AMI
Label w/o AMI
Coronary angiography Percutaneous coronary intervention
12750, 12751, 12752, 12754, 12780, 33110, 33114 33970, 53963, 33974, 33981, 33982, 33983, 33984, 33985, 33986, 33987, 33988, 33989, 3398A, 5396F, 5396G, 5396H, 5396I, 5396J, 5396K, 5396L, 5396M, 5396N, 5396O 53611, 53612, 53613, 53621, 53622, 53623, 5362 53505, 53510, 53511, 53512, 53513, 53521, 53522, 53523, 53524, 53525, 53526, 53527, 53528, 53529, 5352A, 53530, 53531, 53532, 53533, 5357C, 53738, 53739, 53778
‘g’ ‘q’
‘k’ ‘p’
‘c’ ‘n’
‘j’ ‘r’
Surgical coronary revascularisation Non-revascularising invasive procedure requiring invasive coronary status assessment
Author's personal copy Letters to the Editor
139
Survival (%) 100
Stress ECG without successive investigaton
99 98
Stress ECG first
97
Stress ECG – Coronary Angiography sequence
96 95 94
Stress SPECT first
93 92
Stress Echo first
91 90
Elective Coronary Angiography first
89 88 87 0 12 24 36 48 60 72 84 96 108 120 132 144 156 168 180 192 204 216 228 240 252 264 276 288 300 312 324 336 348 360
86
Coronary Angiography during acute myocardial infarction first
Follow-up(days) Fig. 1. Survival curves of individual sub-populations are presented.
Supplementary data to this article can be found online at doi:10.1016/j.ijcard.2011.02.006. References [1] Cooper A, Calvert N, SKinner J, et al. Nice guideline: chest pain of recent onset: assessment of recent onset chest pain or discomfort of suspected cardiac origin; 2010. available at:http://www.nice.org.uk/nicemedia/live/12947/47931/47931.pdf.
[2] Bourdillon PJ. NICE and chest pain diagnosis. Exercise ECG useful in finding coronary artery disease. BMJ 2010;340:c1971. [3] Underwood SR. NICE and chest pain diagnosis. W(h)ither the exercise ECG? BMJ 2010;340:c2387. [4] Timmis A. NICE and chest pain diagnosis. NICE replies. BMJ 2010;340:c2391. [5] Nijjer SS, Eseonu KC, Baker CS. A survey on the latest NICE guidance for diagnosis of chest pain: a challenge to NICE. Int J Cardiol 2010;145:611–3. [6] Shewan LG, Coats AJ. Ethics in the authorship and publishing of scientific articles. Int J Cardiol 2010;144:1–2.
0167-5273/$ – see front matter © 2011 Elsevier Ireland Ltd. All rights reserved. doi:10.1016/j.ijcard.2011.02.006
Heart failure with preserved ejection fraction: A precursor of heart failure with reduced ejection fraction or a distinct syndrome? Daniela Miani a,⁎, Luigi P. Badano a, Paola De Biaggio a, Maria Cecilia Albanese a, Marco Ghidina b, Alessandro Proclemer a, Paolo Fioretti a a b
Division of Cardiology, Department of Cardiopulmonary Science, University Hospital S. Maria della Misericordia, Udine, Italy IRCAB Foundation, Udine, Italy
a r t i c l e
i n f o
Article history: Received 1 February 2011 Accepted 6 February 2011 Available online 3 March 2011 Keywords: Heart failure Survival Preserved left ventricular ejection fraction Outcome Echocardiography
⁎ Corresponding author at: Cardiology Unit, Department of Cardiopulmonary Sciences, University Hospital S. Maria della Misericordia, 33100 Udine, Italy. Fax: +39 0432482353. E-mail address: [email protected] (D. Miani).
To test the hypothesis that heart failure with preserved ejection fraction (HFPEF) may represent an early stage of left ventricular (LV) systolic dysfunction [1,2], we assessed the long-term progression of LV geometry and function in 29 consecutive patients diagnosed with HFPEF [3,4] and compared their long-term survival with that of 102 patients with reduced ejection fraction (HFREF) discharged from our hospital during the same time period. These patients were prospectively followed by our HF outpatient clinic with periodic clinical and echocardiographic controls from December 1999 to December 2007. LV end-diastolic and endsystolic volumes were measured using the biplane area-length method [5]. All measurements were obtained as an average of 3 cardiac cycles. Statistical analyses: continuous variables were summarized as median and first and third quartiles. Expected survival was estimated using the conditional method of Verheul [6]. Differences among diastolic and systolic dysfunction groups were evaluated using a Wilcoxon or a Fisher test, whenever appropriate.
Inpatient events, related International Classification of Diseases (ICD) codes, names and labels used in analysis: Case type
Label
Stabile IHD
‘b’
Angina
‘a’
ICD codes and names I2480 I2500 I2510 I2520 I2530 I2540 I2550 I2560 I2580 I2590 I2000 I2010 I2080 I2090 I2490
Ischemic heart disease, other form Atherosclerotic coronary artery disease Atherosclerotic heart disease Old myocardial infarction Aneurysma cordis Coronary artery aneurysm Ischemic cardiomyopathy Silent myocardial ischemia Chronic ischemic heart disease other form Chronic ischemic heart disease Unstable angina Angina pectoris with proved coronary spasm Angina pectoris, other form Angina pectoris (unspecified) Acute ischemic heart disease
Acute myocardial infarction (AMI) hospitalization
‘i’
I2100 Acute, transmural anterior wall myocardial infarction I2110 Acute, transmural posterior wall myocardial infarction I2120 Acute, transmural myocardial infarction in other location I2130 Acute, transmural myocardial infarction without defined location I2140 Acute, subendocardial myocardial infarction I2190 Acute myocardial infarction (unspecified) I2191 Acute myocardial infarction (unspecified), without Q wave I2200 Anterior wall reinfarction I2210 Posterior wall reinfarction I2280 Reinfarction with other localisation I2290 Reinfarction with not defined localisation I2300 Haemopericardium during acut myocardial infarction I2310 Atrial septal defect developing during the acute phase of myocardial infarction I2320 Ventrucular septal deect developing during the acute phase of myocardial infarction I2330 Free wall rupture during acute myocardial infarction I2340 Papillary cord rupture during acute myocardial infarction I2350 Papillary muscle rupture during acute myocardial infarction I2360 Intracavital thrombus during acute myocardial infarction I2380 Other complications during acute myocardial infarction I2400 Coronary artery thrombosis without myocardial infarction I2410 Dressler syndrome
Inpatient events, related International Classification of Procedures in Medicine (ICPM) codes, names and labels used in analysis: Case type Coronary angiography
Label Label w/ ICPM codes, names w AMI o AMI ‘g’ ‘k’ 12750 Coronary angiography, from other percutaneous entrance 12751 Coronary angiography, from femoral artery entrance 12752 Coronary angiography, from brachial artery entrance 12754 Coronary angiography, from arterectomic entrance 12780 Coronary angiography, from transthoracic entrance 33110 Coronary angiography 33114 Selective coronary angiography
Percutaneous coronary intervention
Surgical coronary revascularisati on
‘q’
‘p’
‘c’
‘j’
33970 53963 33974 33981 33982
Percutaneous coronary angioplasty Coronary angioplasty, per stenosis Coronary stent implantation Coronary stent implantation, right coronary artery Coronary stent implantation, posterior interventricular bracnh 33983 Coronary stent implantation, retroventricular branch 33984 Coronary stent implantation, left anterior descending artery 33985 Coronary stent implantation, diagonal branch 33986 Coronary stent implantation, left circumflex artery 33987 Coronary stent implantation, obtus marginalis branch 33988 Coronary stent implantation, intermedier branch 33989 Coronary stent implantation, left main coronary artery 3398A Coronary stent implantation, bypass branch 5396F Coronary angioplasty, right coronary artery 5396G Coronary angioplasty, left posterior interventricular branch 5396H Coronary angioplasty, retroventricular branch 5396I Coronary angioplasty, left anterior descending artery 5396J Coronary angioplasty, diagonal branch 5396K Coronary angioplasty, left circumflex artery 5396L Coronary angioplasty, obtus marginalis branch 5396M Coronary angioplasty, intermedier branch 5396N Coronary angioplasty, left main coronary artery 5396O Coronary angioplasty, bypass branch 53611 Bypass graft to the right coronary artery 53612 Bypass graft to the left anterior descending artery 53613 Bypass graft to the left circumflex artery 53621 Mammaria impl. on the right coronary artery 53622 Mammaria impl. on the left anterior descending artery 53623 Mammaria impl. on the left circumflex artery 5362A Coronary endoprothesis with open heart surgery
Nonrevascularis ing invasive procedure requiring invasive coronary status assessment
‘n’
‘r’
53505 53510 53511 53512 53513 53521 53522 53523 53524 53525 53526 53527 53528 53529
Transventricular mitral valvuloplasty Opened aortic valvuloplasty Opened mitral valvuloplasty Opened tricuspid valvuloplasty Opened pulmonary valvuloplasty Arteficial aortic valve implantation Arteficial mitral valve implantation Arteficial tricuspid valve implantation Arteficial pulmonal valve implantation Biological aortic valve implantation Biological mitral valve implantation Biological tricuspid valve implantation Biological pulmonal valve implantation Ross operation / pulmonal autograft implantation in aortic position 5352A Aortic homograft implantation 53530 Aortic valvuloplasty 53531 Mitral valvuloplasty 53532 Tricuspid valvuloplasty 53533 Pulmonal valvuloplasty 5357C Aorto-pulmonal homograft implantation 53738 Restitutio endocardii cum transplantatum 53739 Implantatio prothesis endocardii 53778 Pacemaker, defibrillator implantation
Outpatient events, related International Classification of Procedures in Medicine (ICPM) codes, names and labels used in analysis: Case type
Label
Stress ECG
‘e’
Stress Myocardial Perfusion SPECT
‘s’
Stress Echo
‘o’
ICPM codes, names 12620 ECG bicycle exercise 12621 ECG treadmill exercise 12653 ECG tripolar lead system with pharmacological stress 3521D Myocardial perfusion scintigraphy with TlCl (1 timepoint) 35221 Myocardial perfusion scintigraphy with TlCl (2 timepoints) 35223 Myocardial perfusion scintigraphy with Tc-agent 3562A Myocardial perfusion SPECT with Tc-agent 3581A Myocardial perfusion SPECT with TlCl (1 timepoint) 3581B Myocardial perfusion SPECT with TlCl (2 timepoints) 3612M Echocardiography with pharmacological stress
This article appeared in a journal published by Elsevier. The attached copy is furnished to the author for internal non-commercial research and education use, including for instruction at the authors institution and sharing with colleagues. Other uses, including reproduction and distribution, or selling or licensing copies, or posting to personal, institutional or third party websites are prohibited. In most cases authors are permitted to post their version of the article (e.g. in Word or Tex form) to their personal website or institutional repository. Authors requiring further information regarding Elsevier’s archiving and manuscript policies are encouraged to visit: http://www.elsevier.com/authorsrights
Author's personal copy 5012
Letters to the Editor
Regional differences in the utilisation of coronary angiography as initial investigation for the evaluation of patients with suspected coronary artery disease István Kósa a,b,⁎,1, Attila Nemes c,1, Éva Belicza d,1, Ferenc Király a,1, István Vassányi a,1 a
Research & Development Center of Health Informatic, Faculty of Information Technology, University of Pannonia, Veszprém, Hungary Cardiac Rehabilitation Centre of Military Hospital, Balatonfüred, Hungary 2nd Department of Medicine and Cardiology Center, Medical Faculty, AlbertSzent-Györgyi Clinical Center, University of Szeged, Szeged, Hungary d Health Services Management Training Center, Faculty of Health Care, Semmelweis University, Budapest, Hungary b c
a r t i c l e
i n f o
Article history: Received 9 May 2013 Received in revised form 12 July 2013 Accepted 15 July 2013 Available online 8 August 2013 Keywords: Administrative databases Coronary artery disease Coronary angiography Frequency Mortality
Although diagnostic algorithms for the evaluation of patients with suspected coronary artery disease (CAD) are well defined [1–3], the gap between guidelines and daily routine is well known, resulting in considerable variations in the utilization of technologies [4,5]. However, our knowledge is limited regarding the consequences of
this heterogeneity. In the current work we depict a characteristic bias in patient selection in relation to changing diagnostic utilisation. We included 639,139 patients into our analysis, identified by their pseudo social security number from the depersonalised database of the National Institute for Quality- and Organizational Development in Healthcare and Medicines (GyEMSzI), Hungary, for whom diagnostic tests such as stress electrocardiography (ECG), stress echocardiography, stress perfusion single photon emission computed tomography (SPECT) or coronary angiography, were performed to assess symptoms suspect for CAD between 01.01.2004 and 31.12.2008. Survival data were also available from the death registry until 31.12.2009. We classified raw-coded data and merged common coding sequences in a single event, based on the common clinical coding practice. Then we selected patients with stable conditions as those having a new test after a 6 months event-free period. Finally we determined the dominant primary care providers for each ZIP area, based on the provider of the initial diagnostic test, i.e. stress ECG. This allowed us to calculate test frequencies for areas with known population size. To characterize patient subpopulations affected by direct invasive evaluation, we also calculated
Fig. 1. Dominant care provider areas as shaded patches in North-West Hungary. Cities with a population above 10,000 are printed in white.
⁎ Corresponding author at: Cardiac Rehabilitation Centre of Military Hospital, Balatonfüred, Hungary, H-8230, Szabadsag street 5, Balatonfüred, Hungary. Tel.: +36 70 3201192; fax: + 36 87 343434. E-mail address: [email protected] (I. Kósa). 1 These authors take responsibility for all aspects of the reliability and freedom from bias of the data presented and their discussed interpretation.
Author's personal copy Letters to the Editor
5013
Fig. 2. Frequency of direct invasive evaluation at individual cardiological centres.
for each area the mean age, the one year mortality and the age and gender normalized relative mortality of the affected patients. For a more detailed description of the data transformation process, see the Supplement. Our ‘raw-classified event list’ consisted of 2,006,290 events belonging to 639,139 patients. The merging steps reduced the event count to 1,301,135. The 3,860 patients who had missing or invalid ZIP were excluded from the analysis. 121,473 patients of the rest missed the
required event-free period or had an invalid date of birth, but since their permanent residence was known, their event counts could be used to compute the dominant provider of the area. The ZIP area clustering procedure identified 136 dominant primary cardiological centres (Fig. 1). The population size of these areas varied between 255 and 498,328. Out of these 136 centers 85 had a sample size above the predefined limit of 100 tests for direct invasive
Fig. 3. Relative mortality of patients referred for coronary angiography as initial investigation.
Author's personal copy 5014
Letters to the Editor
Fig. 4. Correlation of the frequency of direct invasive investigations and the average age of the population referred for this direct invasive evaluation. Fig. 5. Correlation of the frequency of direct invasive investigation and the relative mortality of the population referred for this direct invasive evaluation.
investigations. These centers covered 88.1% of the population of Hungary with 10,036,878 inhabitants, delivering 89.4% of all performed invasive, and 90.1% of all performed direct invasive investigations. The frequency of coronary angiography ranged between 177.00 and 597.12 per 100,000 inhabitants/year (320.70 ± 84.66) in these 85 centres, while the frequency of coronary angiography as initial investigation was between 24.81 and 151.27 per 100,000 inhabitants/year (67.27 ± 25.87) (Fig. 2). The mean age of the patients in the individual centers affected by the direct invasive patient pathway ranged from 58.35 to 69.50, while the one-year mortality ranged from 1.18% to 15.49%. The age- and gender-adjusted relative mortality was between 0.24 and 2.42 (Fig. 3). BoxPlot analysis identified one center with an outlier value for mortality (15.49%). Since the review of this center's data suggested biased data provision, this center was excluded from further analysis. For the other 84 centres the frequency of invasive investigation correlated well with the frequency of direct invasive investigation (R = 0.64, p b 0.01). The age of patients affected by the direct invasive investigation correlated only with the frequency of direct invasive investigations (R = 0.27, p b 0.05) (Fig. 4.), but not with the frequency of total invasive investigations in the area. Relative mortality decreased both with the increasing frequency of direct invasive investigations and total invasive investigations (Fig. 5) (R = −0.31, p b 0.01 and R = − 0.30, p b 0.01, respectively). The frequencies of the above invasive diagnostic procedures fit well in the row of published European values between 71 and 779 per 100,000 inhabitants [5,6]. Earlier studies were, however, not able to analyze the characteristics of subpopulations. Our sampling method made it possible to identify any point in the evaluation queue of the individual patient, and characterize the affected patient subpopulations. In the current phase of data analysis we focused on the initial test of the patient evaluation process. We selected only cases where patients were referred from the outpatient care directly to invasive diagnostic, without previous hospitalization or noninvasive evaluation during the preceding half year. We found not only a considerable spreading of utilization frequencies of coronary angiography as initial investigation from area to area, but also huge differences in the characteristics of patients referred on this pathway. While the direct referring of high risk patients for coronary angiography is widely accepted, current guidelines suggest the application of noninvasive stress
imaging as initial investigation for the evaluation of patients with moderate risk [1,7,8]. We, however, also found an area from where patients with low mortality risk (1.18%) were referred for direct invasive evaluation. The fact that the frequency of direct invasive tests correlated negatively with the mortality of patients tested, suggests that the method of patient selection is a determinant factor in the formation of procedure utilization disparities. Areas with weaker selection control refer more, but less severe cases for invasive evaluation, while other areas defer most of the low risk patients from the invasive procedure. A limitation of our evaluation method is that it does not account for the follow-up treatment of the patients evaluated invasively. We know, however, that percutaneus revascularization hardly affects survival [9], and that the benefits of surgical revascularization appear only after the first year [10], so this could not explain the observed mortality heterogeneities at one year. Drug treatment could theoretically introduce more prominent deviations if we consider untreated patients versus patients on optimal medical therapy [11], but such great spatial differences within Hungary are hardly expectable, so the effect of this factor should also be very limited. Abbreviations Acute Myocardial Infarction AMI CABG Coronary Artery Bypass Grafting CAD Coronary Artery Disease ECG Electrocardiography GyEMSzI National Institute for Quality- and Organizational Development in Healthcare and Medicines, Hungary ICD International Classification of Diseases ICPM International Classification of Procedures in Medicine SPECT Single Photon Emission Computed Tomography SD Standard Deviation ZIP postal code We gratefully acknowledge the skilled assistance of the staff of the Directorate General of IT and Health System Analysis, National Institute for Quality- and Organizational Development in Healthcare and Medicines (Deputy Director General: György Surján, MD, PhD). The authors of this manuscript have certified that they comply with the Principles of Ethical Publishing in the International Journal of Cardiology.
Author's personal copy Letters to the Editor
References [1] Fox K, Garcia MAA, Ardissino D, et al. Guidelines on the management of stable angina pectoris: executive summary: the Task Force on the Management of Stable Angina Pectoris of the European Society of Cardiology. Eur Heart J 2006;27:1341–81. [2] Kolh P, Wijns W, Danchin N, et al. Guidelines on myocardial revascularization. Eur J Cardiothorac Surg 2010;38(Supplement 1):S1–S52. [3] Smeeth L, Skinner JS, Ashcroft J, Hemingway H, Timmis A. NICE clinical guideline: chest pain of recent onset. Br J Gen Pract 2010;60:607–10. [4] Cook S, Walker A, Hügli O, Togni M, Meier B. Percutaneous coronary interventions in Europe: prevalence, numerical estimates, and projections based on data up to 2004. Clin Res Cardiol 2007;96:375–82. [5] Maier W, Abay M, Cook S, et al. The 2002 European registry of cardiac catheter interventions. Int J Cardiol 2006;113:299–304. [6] Cook S, Walker A, Hügli O, et al. Percutaneous coronary interventions in Europe: prevalence, numerical estimates, and projections based on data up to 2004. Clin Res Cardiol 2007;96:375–82.
5015
[7] Hense HW. Risk factor scoring for coronary heart disease. BMJ 2003;327:1238–9. [8] Califf RM, Armstrong PW, Carver JR, D'Agostino RB, Strauss WE. 27th Bethesda Conference: matching the intensity of risk factor management with the hazard for coronary disease events. Task Force 5. Stratification of patients into high, medium and low risk subgroups for purposes of risk factor management. J Am Coll Cardiol 1996;27:1007–19. [9] Boden WE, O'Rourke RA, Teo KK, et al. Optimal medical therapy with or without pci for stable coronary disease. N Engl J Med 2007;356:1503–16. [10] Hueb W, Lopes N, Gersh BJ, et al. Ten-year follow-up survival of the Medicine, Angioplasty, or Surgery Study (MASS II): a randomized controlled clinical trial of 3 therapeutic strategies for multivessel coronary artery disease. Circulation 2010;122:949–57. [11] Lonn E, Bosch J, Teo KK, et al. The polypill in the prevention of cardiovascular diseases: key concepts, current status, challenges, and future directions. Circulation 2010;122:2078–88.
0167-5273/$ – see front matter © 2013 Elsevier Ireland Ltd. All rights reserved. http://dx.doi.org/10.1016/j.ijcard.2013.07.148
Various morphological types of fragmented ventricular premature beats on 12 lead Holter ECG had positive relationship with LV fibrotic volume on CMR in HCM subjects Koya Ozawa a,1, Nobusada Funabashi a,⁎,1, Hiroyuki Takaoka a, Masae Uehara a, Michiko Daimon a, Marehiko Ueda a, Koji Matsumoto b, Yuji Murakawa c, Yoshio Kobayashi a a b c
Department of Cardiovascular Medicine, Chiba University Graduate School of Medicine, 1-8-1 Inohana, Chuo-ku, Chiba City, Chiba 260-8670, Japan Radiological Department, Chiba University Hospital, 1-8-1 Inohana, Chuo-ku, Chiba City, Chiba 260-8670, Japan The 4th Department of Internal Medicine, Teikyo University School of Medicine, Mizonokuchi Hospital, 3-8-3 Mizonokuchi, Takatsu-ku, Kawasaki 213-8507, Japan
a r t i c l e
i n f o
Article history: Received 11 July 2013 Accepted 16 July 2013 Available online 6 August 2013 Keywords: Various morphological types Fragmented ventricular premature beats 12 lead Holter ECG LV fibrotic volume CMR HCM
Conduction abnormalities within the QRS complex manifest as fragmented QRS waves, which appear as multiple spikes within the QRS wave complex. In patients with coronary artery diseases, fragmented QRS waves have been used as an indicator of non-Q-wave myocardial infarction and are predictors of ventricular arrhythmia [1]. Fragmented QRS waves are important markers for spontaneous ventricular fibrillation in Brugada syndrome [2] and cardiovascular disease [3] and predict the occurrence of high risk syncope and mortality and sudden cardiac death [4].Various morphological types of ventricular premature beats (VPBs) with fragmented QRS waves (fragmented VPCs) are frequently observed in subjects with hypertrophic cardiomyopathy (HCM), but their significance is unknown. In this study, to determine the significance of fragmented VPBs in HCM subjects, we compared the numbers of morphological types of fragmented VPBs and all VPBs with the occurrence of late enhancement (LE) in the left ventricular (LV) myocardium (LVM) on cardiac magnetic resonance (MR) (CMR), which suggests the presence of focal fibrosis.
⁎ Corresponding author. Tel.: +81 43 222 7171x5264. E-mail address: [email protected] (N. Funabashi). 1 These authors contributed equally to this work.
Retrospective analysis was performed in a total of 30 consecutive HCM subjects (21 males, mean age 62 ± 14 years) who underwent CMR (1.5T Intra Achieva Nova Dual, Philips) and a 12-lead Holter ECG (RAC 2103 NIHON KOHDEN) (Fig. 1) within 3 months from July 2007 to April 2012. Patient characteristics and distribution of Maron HCM Types in this study population were represented in Table 1 and Fig. 2, respectively. Detection of LE in the LVM was evaluated by CMR (Fig. 3). Written informed consent was obtained from all patients for all examinations. A fragmented VPB was defined as a VPB with one or more notches in the R or S waves on a routine 12-lead Holter ECG [1,5,6] (Fig. 4). Obvious complete right or left bundle branch block shaped VPBs were excluded from fragmented VPBs in this analysis. The numbers of morphological types of all VPBs and fragmented VPBs were counted automatically, but were manually revised by experienced technologists. After that, an experienced cardiologist blinded to the CMR findings confirmed the results of the printed results. Furthermore from morphology of VPBs, origin sites of LV were estimated using Josephson's criteria [7] (Table 2, Fig. 5) and compared those with LE sites on CMR. The patients were placed supine in a clinical 1.5-T MR imager with 5channel cardiac coils around the chest. All MR images were obtained with ECG gating and during repeated breath-holds. Surface-coil intensity correction was performed for cine MR and late gadolinium enhancement (LGE) MR. Cine MR images were acquired with a steadystate free-precession sequence. After acquiring cine MR images on the 2- and 4-chamber long-axis projections, we obtained short axis cine MR images that encompassed the LV from base to apex [8]. LGE MR was acquired 10 to 15 min after intravenous administration of 0.15 mmol/kg of gadopentetate dimeglumine (Magnevist; Schering AG). An inversion-recovery prepared, T1-weighted, 3D gradient-echo sequence was used to obtain LGE MR in the same planes as cine imaging. The inversion time was adjusted to minimize the signal from normal myocardium in each patient, by using a looklocker sequence to find a null point of normal myocardium. A typical inversion time for LGE MR ranged from 230 to 300 ms.
Diabetes Lifestyle Support with Improved Glycemia Prediction Algorithm Peter GYUK1, Tamas LORINCZ2, Rebaz A.H.KARIM3, Istvan VASSANYI4 Medical Informatics R&D Center University of Pannonia Veszprém, Hungary e-mail:[email protected], [email protected], [email protected],[email protected] Abstract— This paper proposes a combined model to predict the blood glucose level of people with diabetes. Our method consists of two efficient models found in literature and takes nutrition, applied insulin, and initial glucose level into account during the calculations. An extension has been made to these models using various model training methods. Our aim is to help diabetics calculate the insulin need with this efficient algorithm later implemented in a user-friendly software. The tests, that are based on real data, show a significant improvement in the results if the model training methods such as Genetic Algorithm (GA) is used. On the other hand, the numbers reveal the weaknesses of our method, which has to be fixed in the future. During an all-day validation, the prediction error was smaller than 3 mmol/l in 83% of the cases while using GA. Compared to other tests found in literature our model seems to be a good start in predicting glycemia, but needs further improvements. Keywords—Glucose-level tracking; eHealth; Genetic algorithm; Glucose-Insulin system; Glucose absorption; Diabetes mellitus; Outpatient care
I.
INTRODUCTION
Diabetes mellitus is a crucial problem in modern healthcare, since 8% of the population has diabetes in the target age (20-79), according to the recent surveys [1]. Furthermore, the number of people with diabetes may increase by 50% within 2 decades. These numbers remind us of the importance of treating diabetics. Our aim is to provide a tool for them with the help of modern technology and improved prediction algorithms. In case of success, our method can be easily implemented as an add-on to a mobile lifestyle logging application that can be used by many patients to calculate their insulin need. The basic motivation of our efforts is to create a tool that diabetics can use in everyday life to calculate their blood glucose levels. To accomplish this, a reliable method has to be developed to predict the glycemia based on the lifestyle and medication log of an outpatient diabetic. Our previous work [2, 3] showed that the models we chose are capable of a 1-3 hour prediction, but corrections are required to avoid excessive over- and under-estimations. As we reached slightly satisfactory results for the long term (4 or 6 hours) prediction, we started to focus on the model training methods to create better outcomes. There are a lot of methods to be investigated, such as: neural network, fuzzy logic, least square method and genetic algorithm. Some of these have already been applied to the problem of blood glucose
prediction. In the next subsection we give an overview of the current results. A. Literature Overview There are several models available for Blood Glucose Level (BGL) prediction. Most approaches are based on a combination of these models. We review those that include validation on realistic data. The system demonstrated by Stahl et al. [4] consists of three main parts: Glucose Sub-Model (GSM), Insulin SubModel (ISM) and the Glucose/Insulin Interaction Model (GIIM). These three parts are modeled separately using compartment models and linear black-box models [5, 6]. During a 6 months period, input data was collected from a patient diagnosed with Type 1 Diabetes (T1D). Meals, insulin injections and glucose measurements were logged. Researchers had difficulties reaching prediction error smaller than 1 mmol/l in 95% of the cases with 2-hour-ahead prediction. Robertson et al. [7] used Elman’s recurrent Artificial Neural Network (ANN), which predicts BGL based on the history of BGLs, meal intakes and insulin injections. BGL history came from the freeware mathematical diabetes simulator named AIDA. The data set consisted of 28 days and 2688 values. The ANN was trained using all available BGL data for short-term prediction (up to 1 hour). For longterm prediction the ANN was trained with input vector events. Input vector events included 2 meals, 2 short-acting insulin doses, and 2 long-acting insulin doses a day. The maximum error for blood glucose prediction was 0.27 mmol/l for short-term predictions (15, 30, 45 and 60 minutes), 0.2 mmol/l for the 8-hour, and 0.36 mmol/l for the 10-hour predictions, respectively. These are impressive results, however, we must keep in mind that the validation base was a mathematical diabetes simulator data set. In contrast, we used real life measurements of humans. Shanthi et al. [8] carried out the prediction of blood glucose with a simple neural network model, which was trained with the assistance of extracted features. They used a novel feature based prediction algorithm for forecasting the blood glucose values ahead of time. The data set was obtained from diabetic patients in a hospital setting with different insulin therapies using Medtronic Continues Glucose Monitoring System (CGMS). The average errors of this approach are 0.55 mmol/l for the 30 minutes prediction, 0.83 mmol/l for 45, and 1.11 mmol/l for 60 minutes prediction, respectively. These results are promising, but the
validation data was highly controlled, and 50% of data was used for training. In contrast, we used 30% of data for model training with less controlled outpatient data. The sole aim of the Plis et al. [9] study is hypoglycemia prediction. To perform this, they used the Support Vector Regression (SVR) model with physiological features. Instead of tuning parameters, which differ among patients, they used state variables to create features for the SVR model that was individualized for each patient. An extended Kalman filter was run using the training/test points. Input data were collected from 5 T1D patients. The average errors for SVR are 1.25 mmol/l for 30 minutes and 1.99 mmol/l for 60 minutes. There are also other recent approaches not so close to the focus of this paper. Chuah et al. [10], used non-invasive i.e. less reliable blood glucose concentration measurement including healthy volunteers. Seizaburou et al. [11] also used a realistic data set for validation and reached promising results, but without taking meals into account. Liszka [12] used the hybrid Artificial Intelligence technique, which combines the principal component method and the neural networks. However, the authors estimated blood glucose levels only two times a day, while we estimate every 5 minutes. The rest of the paper introduces our model, the validation method, and the results. Section II includes a short overview of our model and presents the model training methods. Section III reviews our testing phases followed by the results and the discussion detailed in Section IV. Section V is a short overview of our software that is developed to support the test process. Finally, Section VI concludes the paper and outlines future works. II.
METHOD
A. Model We created a combined model which reflects the real process happening in our body. Following metabolism, we split the whole procedure in two parts. One of them is insulin absorption, which is simulated with differential equations. The main equations of the chosen model are as follows (3.13.4): 𝑑𝐺 𝑇𝐺𝐻 = −𝐾𝑥𝑔𝑖 𝐺(𝑡)𝐼(𝑡) + 𝑑𝑡 𝑉𝐺
(3.1)
𝑑𝐼 𝑇𝑖𝐺𝑚𝑎𝑥 1 = −𝐾𝑥𝑖 𝐼(𝑡) + 𝑓(𝐺(𝑡 − 𝜏𝐺 )) + 𝑆 (𝑡) (3.2) 𝑑𝑡 𝑉𝐼 𝑉𝐼 𝑡𝑚𝑎𝑥,𝐼 2
Parameter
MODEL SENSIBILITY TEST (RESULTS IN DESCENDING SENSIBILITY ORDER) Change % in results with the given parameter change %
Order
5%
50 %
200 %
𝑉𝑖
5.46
50.23
145.09
1.
𝐾𝑥𝑔𝑖 𝐾𝑥𝑖 𝑇𝑔ℎ 𝑉𝑔
5.27
37.45
71.46
2.
2.25
22.38
82.57
3.
0.27
2.71
10.92
4.
0.26
1.80
3.61
5.
...
The first equation calculates the blood glucose level, depending on insulin absorption, calculated by equation 3.2. This model includes two subcutaneous insulin depots described by equations 3.3 and 3.4. These depots simulate
TABLE I.
...
(3.4)
B. Parameter Identification Algorithms We chose models with large parameter sets to distinguish different patients efficiently. The parameters of the glucose absorption model can be generalized for all patients, because we can presume that these diabetics have healthy digestion. On the other hand, the parameters for the glucose control system are different for each diabetic. Some of these parameters can be measured by an intravenous glucose tolerance test [19], but is too complicated to be made for each person in a realistic outpatient setting. This is the main reason why we need model training methods.
...
𝑑𝑆𝐼 1 =− 𝑆 (𝑡) − 𝑢(𝑡) 𝑑𝑡 𝑡𝑚𝑎𝑥,𝐼 1
The combination of these two models can support diabetics, using subcutaneous insulin injections, no matter if they have Type 1 or Type 2 diabetes [18]. The algorithm properly handles insulin and meal absorption overlaps, as for a longer-time prediction (6-8 hours), the absorption of the insulin and the glucose from food could be in progress during the next meal.
...
(3.3)
Figure 1. The process of absorption from mixed meals
...
𝑑𝑆2 1 1 = 𝑆1 (𝑡) − 𝑆 (𝑡) 𝑑𝑡 𝑡𝑚𝑎𝑥,𝐼 𝑡𝑚𝑎𝑥,𝐼 2
subcutaneous insulin absorption. For further details of this model see [13, 14]. The second part of our combined model describes the glucose absorption from meals [15]. It is a two-compartment model based on mass balance equations. As Figure 1 shows, it divides the digestion into two parts; stomach and intestine. The model takes protein, lipid, monosaccharide, fiber, and starch intake as input, with each one having its own effect during the absorption. This method can deal with mixed meals with components of different Glycemic Indices [16] and takes into account the effect of fiber. Moreover, digestion overlap between two consecutive meals is handled properly. For more details about the model and the parameters see [17].
The running time of the training algorithms depends on the number of the parameters to be trained. As the glucose control model has many variables, a model sensibility test was made to narrow the parameter set (Table I). This means that the training algorithm can run within the same time with fewer parameters to be trained and a wider search range. The results of this test showed that there are 3 parameters which have a significantly larger effect on the results than the others: 𝐾𝑥𝑖 (apparent first-order disappearance rate for insulin), 𝐾𝑥𝑔𝑖 (rate of glucose uptake by insulin-dependent tissues per pmol/l), and 𝑉𝑖 (distribution volume for insulin). These parameters were trained with the methods as described below. For the other parameters, we used an average value suggested by the literature [14]. Two parameter identification algorithms were used. The first one is the Brute Force Algorithm (BFA), which means a full search of parameters in a specific range. BFA analyzes all possible parameter sets within a specific range with the given stepsize. It is not optimal as the stepsize can be decreased ad infinitum, but the returned parameter values are almost perfect. The advantage of this method is its completeness, the disadvantage is the long running time. The other method is the Genetic Algorithm (GA) [20]. GA simulates the process of natural evolution, using the tools of genetics like mutation and crossover. We used an open source library called GAlib [21] in a simple genetic algorithm with one point crossover. The fitness function of the GA was the sum of the differences between measured and estimated blood glucose levels. These training methods themselves have several parameters, henceforth, we performed a test to find the best parameterization. We used 3 data sets including both Type 1 and Type 2 patients. Tables II and III show the results. The best BFA setting was BFA 6, where the stepsize was 0.5, the finding range was 3, and the average running time was 35 seconds. In GA’s case we chose the GA 6 parameterization with the population size of 40, the generation number of 20, the mutation probability of 50%, and the crossover probability of 90%. The high chance of mutation means a more stochastic algorithm. GA 1 was also used during the tests, with the population and the generation of 10, the mutation of 1%, and the crossover of 90%. TABLE II.
BRUTE FORCE ALGORITHM PARAMETER TEST BASED ON TOTAL DIFFERENCE IN MMOL/L
Data set
Base
BFA 1 BFA 2 BFA 3 BFA 4 BFA 5 BFA 6
D1
13.51
9.545
A1
177.94 171.65 161.80 155.41 155.41 159.09 146.26
B1
799.94 417.78 372.76 459.49 375.46 375.29 375.29
TABLE III.
9.33
13.32
10.82
11.68
9.33
GENETIC ALGORITHM PARAMETER TEST BASED ON TOTAL DIFFERENCE IN MMOL/L
Data set
Base
GA 1
GA 2
GA 3
GA 4
GA 5
GA 6
D2
13.97
10.19
9.65
9.67
9.32
9.28
9.18
A2
552.90 430.37 430.99 419.86 417.71 415.57 415.66
B2
600.34 197.84 210.49 208.38 200.37 198.29 196.41
III.
MODEL VALIDATION
The purpose of the validation is to test the prediction power of our algorithm. Accordingly, we used real life data from both type 1 and type 2 diabetics. We expected a significant improvement due to model individualization compared to our previous test using literature parameters [14]. The following subsections review the input data and the validation method. A. Data Sets During the tests we focused on outpatients treated with subcutaneous insulin injections, which means ca. 26% of people with diabetes [22]. We had 7 different data sets of 5 persons, each one including at least 3 days of logging and 12 meals. We had a total of 101 meals and 24 days of input data. As Table IV shows, there were 3 T1D and 4 T2D data sets. Four of the patients used the Medtronic CGMS and one of them (D) used an ordinary blood glucose meter. All 7 logs consist of insulin doses, meals, and blood glucose levels. A professional dietitian calculated the nutriment values for each meal using the hand written logs. Data sets A, B, and C are from the same patient in a controlled experiment, in which the meals were logged rigorously. This patient avoided any sport activities during the monitoring period. In contrast, for the Type 2 patients the meal log may contain inaccurate values as they were cured in hospital to adjust their inordinate glycemia and it wasn't possible to control if they consumed the same meals as offered in the menu. Moreover, sports were also compiled in their log, making the estimations more prone to error because currently the model can’t handle this factor. TABLE IV.
INPUT DATAS
Data set
Type
Age
Insulin
Measure
Meals
Days
A
T1D
21
Apidra
CGMS
15
3
B
T1D
21
Apidra
CGMS
14
3
C
T1D
21
Apidra
CGMS
15
3
D
T2D
62
Humulin R
ordinary
15
6
E
T2D
78
Humalog
CGMS
14
3
F
T2D
61
Humulin R
CGMS
12
3
G
T2D
65
Humulin R
CGMS
16
3
B. Validation Process The validation process consists of 3 phases. The first phase is the study of the model with parameters found in the literature [14]. We made meal wise tests, where the meals were treated as separate tests. This means zero startup blood insulin level and the model starts without any glucose absorption. In this phase, 2 hour, 4 hour, and 6 hour mealwise tests were made to measure the correctness of the model in short-term and in long-term as well. We also made daily tests; one without model restarting and one with model restarting. This means that the estimated blood glucose levels have been set back to the measured value before each meal. This approach is a transition between meal-wise and daily tests, because the insulin and glucose absorption calculations
are continuous, but the blood glucose levels are corrected to avoid stacked errors. In the second phase we performed model training i.e. parameter identification. We made whole day tests with restart using the brute force method and the genetic algorithm according to the model training parameter tests is Tables II and III. In the third phase, we restricted the training data used for the parameter identification to a single day of the log and we used the rest of the log to validate the model with the estimated parameters. We performed all tests mentioned above. Using only a part of the input data for training and the rest for validation avoids over-training and simulates the planned real application of the model in a lifestyle support software tool. IV.
RESULTS
The results were divided in two categories; all patients and controlled data sets. The first one is a simulation closer to reality, while the second highlights the changes in the model as it contains less false data. Test phase 1 (Table V) clearly shows these differences, because in the case of controlled data sets the results were ca. 20% better on average. This improvement for the good of the controlled measurement is caused by the more precise logging. We can also see that longer the time after the meals, the higher the error between measured and estimated blood glucose level. The all-day tests with restarts show a significant improvement in all of the results and the maximum error is also decreased by at least 3 mmol/l. Also, 62% of the errors were within a 3 mmol/l range, which is a promising result for a whole day measurement. This number is even higher (76%) in the case of controlled measurement. TABLE V. TEST PHASE 1: DEFAULT PARAMETERS WITHOUT ANY MODEL TRAINING, AVERAGE VALUES IN MMOL/L (MM) All patients
Meal wise
Whole day
2h
4h
6h
No restart
Restart
Average error
5.05
7.92
9.28
4.2
3.3
Max error
10.62
14.93
17.25
10.34
7.31
< 1mM
34 %
24 %
21 %
22 %
32 %
< 3mM
52 %
43 %
40 %
50 %
62 %
Ratio of error
Controlled (A,B,C)
Meal wise
Whole day
2h
4h
6h
No restart
Restart
Average error
4.26
5.26
5.45
2.5
1.88
Max error
8.1
10.04
11.08
6.94
5.45
< 1mM
34 %
26 %
23 %
29 %
37 %
< 3mM
56 %
51 %
52 %
65 %
76 %
Ratio of error
The “Brute Force” caption in Table VI means the BFA 6 parameterization, described in Section III. “Genetic Algorithm 1” means GA 6 and “Genetic Algorithm 2” means GA 1. With model training, the results show a nearly 25% improvement in average error, but the maximum error almost remained almost the same when model restarting was
applied. With the brute force method, we could reach a ratio 50% for the errors within 1 mmol/l. This means that during a whole day half of the estimated values were in the error range of the measurement devices i.e. 1 mmol/l, so they can be stated as perfect predictions. TABLE VI.
TEST PHASE 2: WHOLE DAY TEST WITH RESTART USING MODEL TRAINING, AVERAGE VALUES IN MM Genetic Genetic Algorithm 1 Algorithm 2
All patients
Brute Force
Average error
1.81
2.18
2.54
Max error
5.83
6.8
7.82
< 1mM
48 %
42 %
40 %
< 3mM
79 %
73 %
71 %
Ratio of error
Genetic Genetic Algorithm 1 Algorithm 2
Controlled (A,B,C)
Brute Force
Average error
1.51
1.64
1.68
Max error
5.24
5.49
5.64
< 1mM
49 %
44 %
43 %
< 3mM
86 %
83 %
83 %
Ratio of error
The reason why model training on whole day tests haven't been made is that we tried to create a real life simulation during test phase 3, where the calculated parameters were tested with the meal wise method. According to our proposal, the future software will make a parameter identification from a few days data flow and will estimate the blood glucose levels after each meal with the calculated parameters. To see how accurate this method is we used BFA 6 and GA 6 to estimate the parameters. Table VII shows the differences in the results to the default parameters. The improvement isn't as significant as in Phase 2, but we can see 5% improvement in average error for GA 6 and an average of 10% for BFA 6. TABLE VII. TEST PHASE 3: REAL USAGE VALIDATION FOR CONTROLLED TESTS, AVERAGE VALUES IN MM Meal wise test (1 h / 2 h / 4 h)
Default parameters
Brute Force
Genetic Algorithm 1
Average error
1.9 / 3.9 / 4.7
1.8 / 2.4 / 4.5
1.8 / 3.7 / 4.7
Max error Ratio of error
4.7 / 7.4 / 8.8
4.2 / 5.1 / 8.3
4.4 / 7.3 / 8.6
< 1mM
53 / 37 / 28 %
54 / 46 / 25 %
55 / 34 / 26 %
< 3mM
79 / 58 / 51 %
80 / 71 / 53 %
80 / 58 / 52 %
A. Discussion of Results The results almost fully confirm our expectations as the model training reached more than 20% improvement in the results. The improvement could be even higher with a longer training sample which was only one day in our current tests. We need more data logs for further tests. Our results are not far from the best results published in the literature. Many other researchers used the Medtronic Guardian CGM system, which indicates that this is a stateof-art device to validate a blood glucose level prediction model. Likewise, in our validation, Stahl et al. [4] had the
same difficulties with the high peak values and they reached 1 mmol/l error in 95% of the cases with 2-hour-ahead prediction. We reached 1 mmol/l error in 46% of the cases with a 2-hour-ahead prediction during controlled tests. We can still improve this result by handling long-term basal insulins, such as Lantus. We experienced that our model can't simulate these long-term insulins properly. The other remarkable result is by Shanthi et al. [8], where the average error was 1.11 mmol/l for 60 minutes prediction, while Plis et al. [9] reached 1.99 mmol/l for 60 minutes. Our best result is 1.8 mmol/l for this period of time with the parameter identification in the controlled measurement. mM 20
15 10 5 0
0
60
120
180
240
300
360 min
Figure 2. Average errors between measured and estimated values in time (solid line – default parameters, dotted line - GA 6 parameters, dashed line - BFA 6 parameters)
As it can be seen on Figure 2 the error is rapidly increasing during the first 2 hours, but the increase slows down between 2 and 6 hours. This shows that the long term
prediction is stable but the difference between the measured and the estimated values is still too large. The graph also shows that the improvements of the model training methods are significant only after the first hour. As we expected, the Brute Force Method gives the best result and the GA is between the BFA and the untrained results. V.
SOFTWARE
A software tool has been designed to support the validation process (Figure 3). The main idea was to provide a useful user interface, which helps us to run the calculations, collect, and process the information. All the data are stored in a relational database. To make the data access faster and consistent for each person participating in the research we chose the PostgreSQL open source database. The patients are organized into groups for the purpose of distinctness by medical experiments. For the implementation of the graphical user interface (GUI) we chose the Qt cross-platform application framework and the C++ programming language Exporting the results to PDF gives us the opportunity to share via e-mail or display on any other devices. With the GUI, the user can select the proper episode of the patient, the start time, and the stop time. The tool lists all the meals, insulins, and measured blood glucose levels. The parameters of each algorithm can be modified before the calculation. After the calculation the results are shown on graphs and tables. The algorithms also provide the optimized values of the parameters. All result are saved in the database for further analysis.
Figure 3. GUI of the software tool: input datas, output graphs and calculated results
VI.
CONCLUSION AND FUTURE WORK
As for the difference between the controlled and all the data, we can state that a more precise logging is needed from the patients. To support this, we plan to create detailed manuals about the important events that should be precisely logged. A clinical study involving 20 diabetic patients will be made in the near future. Extending the 1 day model training period to at least 3 days should bring better results as well. To solve the problems presented in section IV, future research is needed for: improving the currently used model training methods training the model with other parameter identification algorithms extending the model to support physical activity, stress, and weather changes. The final aim is to decrease the average error under 1 mmol/l during the first hour and under 3 mmol/l during the first 4 hours. If the model proves reliable in clinical trials, it will be integrated into the Lavinia [ref.] lifestyle mirror mobile application [ref.] developed at University of Pannonia (http:// www.lavinia.hu/). ACKNOWLEDGEMENT The work presented was supported by the European Union and co-funded by the European Social Fund, project title: “Telemedicine-focused research activities in the field of Mathematics, Informatics and Medical Sciences”, project number: TÁMOP-4.2.2.A-11/1/KONV-2012-0073.
[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15]
REFERENCES [1]
[2]
[3]
[4]
[5]
[6] [7]
L. Guariguata, D.R. Whiting, I. Hambleton, J. Beagley, U. Linnenkamp, J.E. Shaw, Global estimates of diabetes prevalence for 2013 and projections for 2035, Diabetes Research and Clinical Practice, Volume 103, Issue 2, February 2014, ISSN 0168-8227, pp. 137-149. I. Szabo, P. Gyuk, I. Vassanyi. Lifestyle log based blood glucose level predictionfor outpatient care. eTELEMED 2014, The Sixth International Conference oneHealth, Telemedicine, and Social Medicine, 4, 2014. P. Gyuk, I. Szabo, I. Vassanyi, I. Kosa, Levente Kovacs. Combined Model for Diabetes Lifestyle Support. EFMI Special Topic Conference, Budapest, 26-29 April 2014. In L. Stoicu-Tivadar et al (eds): Cross-Border Challenges in Informatics with a Focus on Disease Surveillance and Utilising Big Data, (IOS Press) ISBN 978-1-61499-388-9, DOI 10.3233/978-1-61499-389-6-77, pp. 77-81. Ståhl F., Johansson R., “Mathematical Biosciences Diabetes mellitus modeling and short-term prediction based on blood glucose measurements” Mathematical Biosciences 217 (2009), pp. 101-117. Kajiya F., Kodama S., Abe H. (Eds.), Compartmental Analysis: “Medical Application and Theoretical Background”, Karger, Basel, 1984. Johansson R., System Modeling and Identification, PrenticeHall, Englewood Cliffs, NJ, 1993. G. Robertson, E. Lehmann, W. Sandham, D. Hamilton, Blood Glucose Prediction Using Artificial Neural Networks Trained
[16]
[17]
[18]
[19]
[20] [21] [22]
with the AIDA Diabetes Simulator : A Proof-of-Concept Pilot Study, Journal of Electrical and Computer Engineering Vol. 2011, Article 681786, 11 pages, doi: 10.1155/2011/681786. S. Shanthi, D. Kumar, PREDICTION OF BLOOD GLUCOSE CONCENTRATION AHEAD OF TIME WITH FEATURE BASED, Malaysian Journal of Computer Scince, Vol25 (3), 2012, pp. 136-148. Plis K., Bunescu R., Marling C., Shubrook J., Schwartz F., A Machine Learning Approach to Predicting Blood Glucose Levels for Diabetes Management, Modern Artificial Intelligence for Health Analytics: Papers from the AAAI-14. Zheng-Ming Chuah, Raveendran Paramesran, Kavintheran Thambiratnam, Sin-Chew Poh, A two-level partial least squares system for non-invasive blood glucose concentration prediction” Chemometrics and Intelligent Laboratory Systems 104, 2010, pp. 347-351. Seizaburou A., Masaya Y., Tadashi I., System and Method for Predicting Blood Glucose Level, Patent No. 5971922 Oct. 26, 1999. Liszka-Hackzell, J J, Prediction of blood glucose levels in diabetic patients using a hybrid AI technique, Computers and biomedical research 32, 1999, pp. 132-144. Palumbo P., Pepe P., Panunzi S., De Gaetano A. (2011), Glucose control by subcutaneous insulin administration: a DDE modelling approach, in Proc. 18th IFAC World Congress, Milan, Italy, 2011, pp. 1471-1476. P. Palumbo, P. Pepe, S. Panunzi, A. De Gaetano, Observerbased glucose control via subcutaneous insulin administration, Biological and Medical Systems, Vol. 8, 2007, pp. 107-112. G. Eason, B. Noble, and I. N. Sneddon, “On certain integrals of Lipschitz-Hankel type involving products of Bessel functions,” Phil. Trans. Roy. Soc. London, vol. A247, April 1955, pp. 529–551. Arleth T., Andreassen S., Orsini-Federiri M., Timi A., MassiBenedetti M., A model of glucose absorption from mixed meals, IFAC, 2000. Jenkins DJA, Wolever TMS, Taylor RH, Barker H, Fielder H, Baldwin JM, Bowling AC, Newman HC, Jenkins AL. Glycemic index of foods: a physiological basis fo carbohydrates exchange. Am J Clin Nutr 1981: 34; pp. 362366. Arleth T., Andreassen S., Orsini-Federici M., Timi A., MassiBenedetti M., Optimisation of a model of glucose absorption from mixed meals., 2005. American Diabetes Association. "Diagnosis and Classification of Diabetes Mellitus." Diabetes Care 27.90001 (2004): 5S-10. Panunzi S., De Gaetano A., Mingrone G. (2010), Advantages of the single delay model for the assessmentof insulin sensitivity from the intravenous glucose tolerancetest,Theor. Biol. Med. Model., 7, pp. 1-20. S.N. Sivanandam and S.N. Deepa. Introduction to genetic algorithms. Springer, 2008. M. Wall, GAlib: A C++ Library of Genetic Algorithm Components, August 1996, http://lancet.mit.edu/ga/. Centers for Disease Control and Prevention. National diabetes fact sheet: national estimates and general information on diabetes and prediabetes in the United States, 2011. Atlanta, GA: U.S. Department of Health and Human Sersvices, Centers for Disease Control and Prevention, 2011.
í
!
"
#
$
%
&
'
(
)
&
6
F
G
L
H
M
T
N
P
Q
V
T
Y
N
X
X
U
T
Q
T
V
N
R
R
Q
R
Q
P
Q
N
S
P
^
R
X
M
M
Q
Z
O
O
P
X
O
T
N
O
Z
W
R
M
X
T
N
M
_
b
h
Q
W
`
W
Q
h
Q
Q
N
_
T
P
V
W
V
%
)
!
N
X
*
$
+
!
!
0
,
8
O
X
M
-
9
^
T
Q
_
N
^
X
^
R
\
)
:
P
Q
.
;
U
O
/
*
<
M
[
^
M
O
'
!
Q
&
"
#
M
=
0
4
O
T
)
"
N
W
X
$
!
M
M
W
X
W
T
-
0
#
*
]
Y
M
1
>
!
?
W
T
W
^
#
@
!
A
)
;
V
g
S
M
U
Q
M
S
T
"
-
;
\
N
2
ð
ñ
ò
)
>
X
ï
,
B
S
^
X
Q
T
ó
ô
ï
ö
!
1
õ
!
D
ò
3
C
U
M
X
\
N
V
P
X
.
÷
ò
&
õ
4
ø
ù
0
÷
!
ò
$
ú
5
ó
!
ô
$
ó
"
ò
û
ó
ó
ü
ý
÷
û
ö
ó
ï
þ
ÿ
í
!
E
T
P
R
W
P
T
P
]
[
Y
U
Q
U
W
M
W
Q
k
X
X
S
U
Q
\
N
U
U
Q
U
Q
N
Z
X
P
Q
Q
T
S
O
W
T
X
Q
^
S
Q
N
O
\
X
b
N
M
U
]
T
P
M
N
c
U
U
X
X
Q
P
Y
M
P
^
e
f
V
M
X
N
^
S
M
P
R
M
Y
U
X
N
Q
^
U
Q
[
Y
]
X
b
b
b
Q
M
X
N
[
T
M
Q
P
S
R
_
b
P
O
_
N
M
T
X
U
Z
_
N
O
Q
Q
Y
Q
W
T
N
U
^
X
M
Q
V
S
[
_
P
]
S
M
Q
T
X
N
W
R
Q
O
N
X
]
X
W
U
N
Q
T
U
N
O
O
S
P
N
Z
X
S
T
Z
M
_
M
[
[
X
Q
]
M
g
a
W
Z
N
\
`
c
N
M
d
W
M
O
Z
X
M
R
S
M
c
`
O
W
P
S
N
M
W
N
N
M
Z
O
M
N
\
Z
M
W
\
W
W
X
Q
Z
X
Z
[
P
^
T
^
N
M
O
U
M
N
O
Z
X
N
X
_
S
i
Q
^
N
W
N
P
O
R
M
P
j
Q
N
_
Y
`
X
T
N
N
W
P
T
O
\
^
M
Q
W
\
Y
P
]
P
O
U
T
P
Z
O
N
Q
T
V
M
R
M
T
Q
P
Q
X
T
]
O
N
X
W
Q
M
N
Z
_
R
Z
X
P
Q
S
Q
M
X
R
N
S
N
X
W
U
\
_
Q
S
N
M
M
Y
X
R
Z
M
X
M
l
N
P
O
^
P
P
Q
\
O
M
N
R
X
d
g
Y
N
W
`
T
O
Q
R
[
_
^
^
P
\
M
M
P
U
Q
W
\
N
N
W
W
U
c
O
W
M
X
X
M
M
U
U
^
U
Q
[
O
P
Y
_
M
O
P
M
Y
V
V
O
T
V
W
U
O
P
Q
Q
M
U
M
M
X
U
N
U
\
]
P
N
[
^
d
M
M
V
M
W
W
`
]
T
[
Y
X
M
W
M
N
N
N
O
W
b
Q
O
T
d
Q
S
N
S
X
Y
W
W
M
Z
`
Z
T
\
N
M
N
^
X
V
]
O
P
O
M
Q
[
\
[
P
U
Z
Z
P
P
T
b
M
Z
T
V
N
N
Q
Q
U
U
N
X
N
U
N
P
X
Q
T
O
Q
U
Z
O
M
Q
]
Q
[
X
M
M
N
b
O
P
\
L
N
`
Z
T
P
X
\
N
W
Z
V
\
M
Z
Q
X
Z
O
T
M
P
T
V
M
U
N
Y
N
O
Y
X
X
T
X
X
O
P
P
W
g
W
T
T
Y
M
M
M
Q
W
U
_
W
V
S
T
W
N
Q
_
P
W
[
U
Q
M
M
M
O
\
W
Z
O
M
X
Z
W
Z
N
S
W
M
N
M
R
X
Z
Z
T
T
P
Q
U
Q
O
_
T
X
b
X
S
M
X
7
ï
I
N
\
N
M
K
R
X
N
P
b
R
P
_
Q
N
F
_
M
N
Q
T
X
J
P
O
W
^
U
O
M
P
W
I
î
T
W
W
O
T
g
V
Q
X
\
l
Q
X
Z
\
M
P
W
Z
N
U
M
J n
o
d
p
M
U
t
M
_
M
^
N
U
W
W
T
[
W
M
b
T
^
Y
m
g
T
O
P
\
M
N
O
X
O
N
X
M
W
M
P
W
[
X
T
O
W
X
T
X
O
N
Z
S
\
Y
W
Q
Z
N
U
N
Z
X
Q
]
Q
X
Y
Z
T
_
b
Q
P
M
^
g
Q
X
\
_
N
P
[
N
U
U
T
R
R
Q
P
R
`
m
c
_
T
b
Q
U
M
Z
T
_
S
Q
P
R
R
T
U
w
P
T
Q
z
X
Q
O
v
X
[
x
P
c
M
}
U
Y
|
X
M
{
V
Q
U
z
c
M
y
P
Q
P
H
x
s
Z
Q
w
^
v
N
r
u
~
q
W
§
¡
¡
¡
¡
¢
¢
£
¤
¥
¦
ª
Ï
Ð
Ñ
á
Ò
Ó
ì
Õ
Ô
Ï
Õ
Ø
Ö
Ô
Ø
Ó
á
Ù
Ü
Ø
Ú
Ø
Ý
Û
Þ
à
à
Ý
Ø
Ñ
Ø
Ð
Ý
ß
à
à
Ò
Ñ
Ò
á
Ñ
â
Õ
ã
Ò
â
ä
ã
å
Ñ
à
ä
Ò
å
Õ
Ð
Þ
Ý
Ý
Ë
Ò
Ó
Þ
Õ
à
à
æ
Ð
Õ
Ñ
à
Õ
Ô
Ð
è
Ô
Ò
Þ
Ó
ç
ã
Ô
æ
â
å
è
±
´
¿
¼
½
®
¢
¤
À
º
°
¤
¬
¡
¸
»
´
·
º
¸
¹
³
¸
±
·
°
©
µ
·
®
±
°
¶
µ
³
²
°
¯
·
Ý
é
Í
å
Ó
á
Õ
æ
Ü
Õ
Ò
Õ
±
Þ
å
¸
³
×
Õ
Æ
Ø
·
á
°
â
Æ
Ô
°
ã
Ñ
º
È
Ò
¬
×
º
ä
Ô
¸
Ø
°
Ö
â
Ã
¬
Ò
¯
Ñ
±
â
Õ
®
»
å
Ü
Ò
æ
¬
Ì
ä
Õ
¬
¸
¢
Å
Ñ
ª
Ø
æ
®
º
æ
ç
Þ
·
¬
¸
³
Þ
Ô
¬
Ý
Ð
Ý
Ø
Ä
»
Ø
»
Â
¸
°
à
à
Ä
¢
½
Ò
Ó
¬
Ê
â
Ø
Õ
Þ
¡
¶
·
Ò
Ò
µ
±
·
Ê
Ã
Ð
ã
É
Ý
¾
°
®
ã
¸
â
¶
Þ
á
¨
Ü
ß
«
¢
Ó
ª
¤
¤
Õ
×
Ð
Ò
ì
¡
Ô
Î
ê
¢
Á
¤
¡
~
æ
Ô
Ð
Ó
Ä
·
Ç
Ð
°
â
Å
á
Â
±
Ç
¬
·
Ç
º
Æ
°
Å
¼
®
±
°
®
´
æ
Ø
ë
ÿ
ñ
!
ý
"
#
¢
z
Â
´
º
¬
»
·
±
·
·
´
¬
§
¢
°
Æ
·
¸
¸
°
È
¿
¥
¤
¼
»
´
·
®
Ä
¸
Å
¼
±
·
±
·
Ç
¬
®
¬
¸
±
·
µ
»
°
·
¬
¬
¸
±
¼
·
º
º
·
¬
Ã
º
®
Æ
¿
¸
Å
°
´
³
¸
®
°
Ä
·
°
Æ
Å
°
®
Å
Ç
Å
°
¬
°
Æ
·
º
²
¬
¬
Ä
¬
¶
»
¸
®
Ã
±
¸
°
Ã
®
¢
º
´
°
®
Æ
½
·
·
Ã
Å
¸
È
°
º
°
·
®
º
³
¬
®
°
®
º
Ã
¤
º
¤
µ
±
½
°
©
Á
»
½
¼
»
°
º
´
Ã
·
³
²
³
®
¡
§
·
®
¸
¼
Å
~
È
±
±
´
·
³
Ã
®
³
º
¸
°
¬
Æ
¯
Å
¢
ª
°
Ä
¸
°
³
Å
®
¸
º
´
Â
Å
¼
º
¬
¬
§
§
¥
¤
¡
{
í
ÿ
x
~
&
ü
ÿ
î
§
ü
î
î
%
ü
%
½
ü
u
¨
¡
%
$
£
¡
¡
¥
¦
¡
½
¢
°
¸
Ä
®
·
Æ
·
´
º
Ã
Å
¸
Å
¥
°
¸
¯
°
±
¸
Ç
°
¬
µ
°
º
°
°
Ã
°
Ã
·
°
º
°
±
¸
Ó
Ð
ÿ
Ð
ä
Ô
×
Õ
Ô
Ó
Ð
Ô
$
Ò
é
Õ
â
Þ
Ø
Ü
á
Ø
Ñ
Ø
Ý
Þ
Õ
ß
à
Ô
ë
à
Ó
Ý
Ó
Ñ
Ò
Ð
á
Ó
â
â
Ü
ã
Ò
Ø
Ø
Ò
Ý
ã
ß
Ñ
à
à
Õ
â
ä
Ò
å
à
â
Ñ
Þ
Ð
Ø
Ò
Ó
Ô
×
Õ
à
Ô
Ð
Ð
Ý
â
Ü
Ý
Ö
Þ
à
æ
Õ
Ø
Ð
Ô
Ñ
Þ
ç
½
Ô
æ
è
ä
Ò
å
â
á
Ò
×
Ô
â
Ø
Ó
Ð
ä
Õ
Ó
ã
Ò
Å
·
Ä
°
Ã
¸
²
¼
®
µ
°
º
´
¬
®
³
¸
í
'
Q
R
[
O
M
e
(
`
^
N
Q
U
Y
^
Q
M
X
N
O
Y
U
T
R
W
[
_
_
N
O
)
Y
U
M
V
*
X
N
P
^
X
\
M
_
M
N
W
[
O
M
O
Y
^
V
T
T
^
S
Y
O
N
_
î
Q
ï
^
ð
ñ
V
T
ò
ï
O
ó
ô
N
ò
T
õ
P
ï
M
ö
m
÷
ò
_
õ
T
ø
P
ù
X
÷
\
ò
S
ú
M
ó
O
Q
ô
T
ó
ò
û
)
^
ó
O
Q
ó
R
ü
\
ý
÷
*
X
û
Q
ö
P
ó
X
ï
\
þ
ÿ
M
í
[
W
M
O
l
Q
P
X
M
O
T
V
V
N
X
P
[
\
X
¡
M
Q
Á
¡
¡
¢
W
`
*
¡
X
M
M
P
\
O
[
M
X
R
X
V
P
T
Q
T
X
V
Z
P
M
X
[
P
\
M
T
X
Q
P
O
_
Q
X
N
Y
R
[
^
O
P
M
M
Y
R
P
R
R
O
T
M
U
M
U
N
P
M
X
M
\
T
M
X
X
\
P
^
X
M
P
N
P
\
X
N
g
T
M
V
Z
R
N
W
Á
¢
b
N
N
^
S
\
M
M
Z
M
\
[
V
X
W
X
V
P
)
X
N
T
Á
X
\
X
¡
W
W
P
^
^
Á
Q
Q
M
P
g
Z
N
M
k
O
Q
W
M
X
X
g
P
P
.
[
T
M
T
\
M
W
\
_
X
§
W
N
^
¡
N
M
N
\
_
M
X
_
O
V
[
S
Q
W
W
^
R
X
M
O
T
N
\
Y
U
X
U
M
O
U
\
N
N
Q
X
d
b
O
`
P
P
N
T
M
S
Q
M
X
O
N
W
P
Z
R
O
Q
N
U
[
X
S
V
^
S
T
P
N
_
N
^
O
c
Q
T
(
T
V
,
O
M
^
\
P
X
+
N
P
N
M
Q
\
P
Q
X
Q
_
W
]
M
M
X
\
Z
N
Q
L
^
N
¡
M
M
M
\
R
O
X
R
X
V
T
P
-
¥
M
¡
T
U
Q
M
M
O
Z
£
~
½
¥
~
~
£
/
u
/
u
1
t
°
·
Æ
Æ
4
¬
3
¬
¸
Ä
½
2
µ
º
°
±
°
·
±
°
±
³
|
}
v
z
w
<
³
7
v
0
{
;
ª
´
:
¡
{
9
¥
w
8
°
7
¢
1
6
¼
¦
x
5
³
0
|
2
Á
2
¡
Å
y
Ä
²
¬
Ã
°
º
.
½
¬
Æ
¬
°
¶
Ã
¡
²
º
¬
°
°
±
·
¸
Æ
»
®
Å
±
¬
°
º
µ
Å
A
®
¬
@
?
>
C
>
=
Å
³
?
½
Â
A
@
>
B
C
¶
¤
Ä
°
º
Ç
¦
·
¢
°
·
Ã
¡
Å
¤
°
·
Æ
Æ
¬
³
¬
Ä
¼
¸
ÿ
$
í
ñ
!
ý
'
Q
"
R
#
[
O
$
%
D
M
ü
`
d
%
\
%
ü
M
^
Q
W
î
X
O
î
Q
ü
b
î
[
ÿ
X
Q
ü
T
&
P
ÿ
T
E
V
í
[
N
P
X
O
/
u
½
ÿ
$
¿
Z
N
U
)
M
b
N
O
*
W
H
P
[
_
b
M
O
T
V
Z
N
W
M
W
c
¦
¤
¦
¢
¥
£
W
X
V
W
M
M
L
W
N
`
Z
G
P
V
Q
T
O
M
T
R
O
N
O
X
P
M
M
M
Z
\
O
X
M
W
S
Â
g
^
T
M
\
X
W
N
W
U
Q
k
[
N
_
U
[
.
¢
~
~
¡
¥
¦
ª
¤
¿
¦
~
½
¦
¢
£
¿
¥
¥
½
½
¥
9
K
;
¥
O
N
Z
X
H
;
¿
M
N
P
*
M
7
M
T
P
Q
K
F
Q
U
O
^
T
M
v
X
\
X
L
M
T
\
^
d
~
K
`
)
W
M
5
¡
;
2
K
O
U
¤
T
N
¡
O
Z
¥
O
W
½
M
X
5
Y
\
¥
X
R
9
J
¢
;
¦
ÿ
¦
Á
2
2
Á
.
1
4
;
I
|
¦
¥
½
¡
¡
Q
Q
í
P
u
}
1
»
\
³
°
³
°
Q
Ä
»
±
°
°
³
w
U
º
¬
µ
}
Ã
´
¬
¶
º
°
Y
[
^
z
§
ñ
ò
ï
ó
ð
ô
ò
õ
ï
ö
÷
ò
õ
ø
ù
÷
ò
ú
ó
ô
ó
ò
û
ó
ó
ü
ý
÷
û
ö
ó
ï
þ
ÿ
í
S
y
R
y
ï
x
|
T
x
|
w
°
±
³
°
Q
{
µ
w
1
Ä
{
°
X
·
y
[
Z
w
z
0
Y
°
]
®
¬
v
R
³
z
º
°
¬
y
|
X
·
w
±
S
}
0
}
w
z
î
_
`
a
.
V
~
¡
¡
È
¡
È
W
À
¨
¡
Á
½
V
Â
£
l
b
N
M
O
i
+
c
2
!
c
`
"
)
`
4
d
c
%
9
Q
O
f
e
W
f
Z
\
g
M
j
P
[
b
N
k
P
[
D
_
l
,
c
)
g
*
H
l
b
[
T
M
[
N
X
O
c
8
M
O
O
M
N
U
U
O
U
W
N
Q
g
r
`
M
M
8
O
"
)
b
`
`
c
f
e
m
f
f
l
f
g
e
`
M
U
V
m
_
T
P
Q
X
T
O
Q
P
R
_
N
Y
b
M
P
M
Z
M
W
W
N
O
Y
V
T
O
W
[
Z
Z
M
W
W
V
[
U
g
M
Q
R
\
X
Z
T
P
X
O
T
U
h
`
!
2
)
&
"
`
`
.
d
&
`
'
e
g
f
e
2
e
$
`
!
`
#
d
\
D
c
D
e
`
w
`
-
$
#
c
"
g
e
&
!
W
X
)
$
u
!
-
V
[
O
Q
q
\
X
Q
P
T
[
Q
V
P
R
U
S
N
U
Q
T
S
p
^
O
s
W
u
X
*
e
O
f
N
P
e
M
f
e
M
T
D
X
9
^
p
m
)
2
P
U
&
'
"
O
M
`
)
M
e
`
f
#
M
e
-
f
e
%
M
q
#
T
M
W
X
V
)
Y
U
g
e
T
*
M
M
H
M
X
x
o
P
_
O
k
Y
Q
[
O
H
N
^
W
Q
T
[
W
l
W
L
W
W
D
Q
X
j
P
m
^
\
c
T
f
e
R
e
Z
D
P
Q
e
W
X
M
H
W
H
M
e
P
(
e
t
c
X
W
N
M
U
Y
R
V
M
m
W
X
M
H
W
_
_
f
T
N
D
P
Q
m
X
X
T
Q
O
+
Z
q
e
M
]
Q
M
g
T
V
X
\
M
L
Q
X
M
O
N
X
[
O
M
9
D
Q
P
R
N
W
N
]
Q
X
N
U
Z
T
_
S
T
P
M
P
X
T
V
`
M
N
W
M
W
T
V
Z
Q
]
Q
U
F
Q
N
X
Q
T
P
9
1
!
#
!
)
"
-
2
1
!
4
&
"
$
#
5
.
-
)
.
h
c
$
!
,
v
`
(
`
$
c
U
q
m
h
!
_
T
p
c
[
O
D
c
Z
3
N
X
e
j
c
U
Q
P
D
]
)
P
T
N
H
-
M
Z
X
e
M
\
X
e
Y
&
"
Z
\
q
e
c
!
W
R
`
D
,
P
t
c
O
Q
`
0
Q
M
%
j
c
D
d
W
7
R
!
g
T
c
T
P
2
c
U
&
`
X
a
X
.
(
`
[
$
[
W
&
N
'
`
V
N
,
n
X
&
W
b
m
"
o
c
.
d
M
N
c
Q
`
)
c
Z
)
W
M
Z
n
`
"
U
[
5
d
L
&
W
c
f
l
O
r
b
e
`
g
!
u
D
f
s
`
0
!
"
N
-
U
)
Q
^
Q
3
X
Y
!
T
$
V
!
$
N
V
-
D
T
#
T
#
^
&
V
-
)
O
$
E
M
&
[
M
P
)
Z
E
Y
f
e
u
[
*
f
M
W
s
X
T
Q
U
T
[
P
P
_
N
Q
H
M
O
u
M
N
f
c
P
j
W
^
W
N
[
V
H
M
T
T
q
e
^
^
t
c
N
Q
R
N
M
O
Y
Q
H
W
P
g
e
N
l
f
W
m
g
e
\
T
f
f
O
X
m
X
M
O
_
O
M
Z
N
U
U
l
+
[
W
Q
P
O
p
c
M
]
y
c
Q
M
g
O
`
/
W
N
$
P
!
^
"
n
c
)
z
c
$
&
N
)
O
X
|
.
]
Q
&
R
8
W
"
M
P
)
{
c
.
&
`
'
D
c
0
!
q
,
g
e
-
'
`
)
.
[
P
'
Z
&
"
X
Q
0
T
P
)
$
N
U
-
Q
#
X
Q
M
D
9
W
N
q
P
^
g
e
Q
(
P
[
S
[
k
R
X
u
_
D
)
M
u
X
*
H
\
T
x
^
W
g
h
V
m
x
T
O
l
O
M
Z
T
O
^
Q
P
R
V
T
T
^
Q
P
X
N
H
M
(
W
Y
W
X
M
_
N
X
Q
Z
`
l
e
M
]
Q
Z
N
W
M
U
p
c
V
m
(
_
T
c
P
M
Q
X
X
T
N
O
Q
U
P
`
D
c
R
Q
q
q
P
u
_
,
`
N
P
M
N
R
W
Q
Q
P
R
P
R
X
c
V
Y
S
M
N
W
Q
D
M
b
Q
^
Q
U
N
Q
X
b
Y
M
c
X
N
M
P
W
^
`
N
5
Z
&
Z
$
M
S
!
X
0
N
4
b
&
Q
"
U
Q
)
X
"
Y
T
%
V
5
.
N
P
Q
-
)
P
X
M
O
i
.
]
"
M
)
P
.
X
Q
T
P
D
#
[
f
)
W
D
Q
q
P
R
q
u
S
*
M
g
O
f
W
T
x
P
N
l
m
q
U
^
f
Q
R
Q
X
N
U
N
W
W
Q
W
X
N
P
X
m
b
N
W
M
^
`
l
e
]
M
X
M
Z
s
N
W
W
N
N
R
X
P
M
T
U
c
U
`
&
j
X
X
D
q
x
Q
T
h
m
N
O
d
c
c
e
M
^
`
x
`
M
j
9
c
D
t
`
T
Q
'
O
`
L
O
Y
&
P
P
c
P
5
o
Y
T
r
j
!
`
D
c
`
N
0
e
`
U
W
&
e
U
N
`
u
Q
-
\
}
P
)
N
c
`
D
t
M
q
+
c
,
b
q
D
X
.
]
M
6
Q
T
$
O
b
7
O
N
S
|
!
c
T
_
2
U
`
P
O
8
!
X
T
&
P
X
T
m
P
X
2
N
u
Q
W
$
\
D
c
V
'
c
`
7
M
T
.
g
]
P
)
e
O
W
q
"
M
Q
"
D
c
!
j
N
&
U
D
O
X
M
D
U
M
T
Q
!
R
a
X
R
0
P
f
W
N
!
M
T
t
W
W
]
O
M
"
W
U
M
X
M
[
+
^
b
T
O
P
_
_
T
)
N
M
T
V
-
Q
,
X
Q
q
Q
N
,
_
D
O
W
!
X
Q
g
b
P
R
D
P
M
#
M
N
W
g
#
S
d
X
P
.
c
M
Q
)
W
e
N
Q
-
O
[
e
T
N
U
Q
-
W
P
M
M
&
P
c
m
p
\
)
X
$
N
W
V
R
U
Q
N
W
^
\
N
N
R
X
T
^
[
H
W
s
`
M
X
P
T
F
Q
L
U
&
Q
P
M
M
Q
_
\
U
X
N
Y
U
e
X
M
e
`
3
Q
H
M
O
j
P
g
_
N
c
M
M
[
R
\
D
T
R
f
R
Q
D
a
)
X
f
w
T
j
P
W
W
D
$
P
N
H
M
N
g
4
u
!
Q
P
q
-
P
q
e
^
g
m
#
_
[
N
f
!
W
W
R
q
!
U
c
N
Y
S
"
M
8
P
t
U
c
$
X
c
S
0
N
N
W
P
R
Q
D
W
D
p
Z
M
M
`
N
"
P
g
e
&
h
Y
N
e
(
D
m
O
H
W
-
^
T
f
_
m
9
:
~
F
Q
e
$
M
l
e
^
u
2
/
$
9
`
D
q
e
e
]
T
U
D
`
l
D
ÿ
$
î
Hindawi Publishing Corporation Journal of Healthcare Engineering Volume 2016, Article ID 5136705, 13 pages http://dx.doi.org/10.1155/2016/5136705
Research Article Stress Detection Using Low Cost Heart Rate Sensors Mario Salai, István Vassányi, and István Kósa Medical Informatics R&D Centre, University of Pannonia, Egyetem Utca 10, Veszpr´em 8200, Hungary Correspondence should be addressed to Istv´an Vass´anyi; [email protected] Received 11 March 2016; Accepted 5 May 2016 Academic Editor: Valentina Camomilla Copyright © 2016 Mario Salai et al. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. The automated detection of stress is a central problem for ambient assisted living solutions. The paper presents the concepts and results of two studies targeted at stress detection with a low cost heart rate sensor, a chest belt. In the device validation study (𝑛 = 5), we compared heart rate data and other features from the belt to those measured by a gold standard device to assess the reliability of the sensor. With simple synchronization and data cleaning algorithm, we were able to select highly (>97%) correlated, low average error (2.2%) data segments of considerable length from the chest data for further processing. The protocol for the clinical study (𝑛 = 46) included a relax phase followed by a phase with provoked mental stress, 10 minutes each. We developed a simple method for the detection of the stress using only three time-domain features of the heart rate signal. The method produced accuracy of 74.6%, sensitivity of 75.0%, and specificity of 74.2%, which is impressive compared to the performance of two state-of-the-art methods run on the same data. Since the proposed method uses only time-domain features, it can be efficiently implemented on mobile devices.
1. Introduction Stress is commonly defined as a feeling of strain and pressure [1]. There is evidence that stress is linked with many diseases, playing a crucial role in the development of cardiovascular diseases [2], diabetes [3], or asthma [4], and it also significantly influences the later course of these diseases. Stress is related to life style; therefore, especially for mobile automated lifestyle counseling and analysis services, the need arises to identify stress automatically during daytime, using physiological data from various sensors. If stress could be reliably and automatically identified, this could directly help users manage stress situations, and it could also be used in medical intelligence applications, for example, in refining blood glucose predictions for diabetics during daytime under influence of stress. However, the available methods for automated stress detection based on low price, ubiquitous sensors, are yet immature. Telemonitoring and self-management systems [5– 9] extend the horizons of traditional health care using only point of care measurement data, but the proper interpretation and reliability of the results depend on the reliability of the measured data and the sensor itself. The two crucial questions related to this problem are as follows:
(i) Whether low price physiological sensors are reliable enough compared to “gold standard” devices accepted by and used in clinical practice. (ii) Which sensors and algorithms can provide a reliable method for stress detection, at an affordable price and minimal user interaction. This paper describes our efforts and results in answering these questions. The rest of this paper is organized as follows. Section 1 gives an overview on applicable technologies and related research. In Section 2, we describe our methods for a small scale device validation and the methods used in the main clinical study that compared heart rate variability (HRV) features between relaxation and mental stress periods. We also present here a simple stress detection algorithm. Section 3 presents the measurement results of the two studies, with the latter compared with two state-of-the-art algorithms. We conclude the strengths and weaknesses of this study and the newly designed stress detection algorithm in Section 4. The most commonly used physiological markers of stress are as follows: (i) Galvanic skin response (GSR): using changes in skin conductivity. During stress, resistance of skin drops due to increased secretion in sweating glands [10].
2
Journal of Healthcare Engineering (ii) Electromyogram (EMG): measuring electrical activity of the muscles. Stress causes differences in the contraction of the muscles which can be used to identify stress [11, 12]. (iii) Skin temperature: changes in temperature of the skin are related to the stress level [13]. (iv) Electrical activity of the heart: the most commonly used stress marker parameters derived from the electrocardiograph (ECG) are the heart rate (HR) and the heart rate variability (HRV) [14].
Stress can also be detected using other, less common markers like accelerometer [15], key stroke dynamics [16], or blinking [17]. It is also common to use a combination of several markers at the expense of an increased system cost and user involvement. Fernandes et al. used GSR and blood pressure (BP) markers [18] for determining stress. Sun et al. describe mental stress detection using combined data from ECG, GSR, and accelerometer [19]. De Santos Sierra et al. in [20] used GSR and HR. Rigas et al. used ECG, GSR, and respiration for detecting stress while driving [21]. Wijsman et al. used ECG, respiration, GSR, and EMG of trapezius muscles for mental stress detection [22]. Riera et al. combined EEG and EMG markers [23]. Singh and Queyam used GSR, EMG, respiration, and HR [24] for detecting stress during driving. Pupil diameter, ECG, and photoplethysmogram were used as markers by Mokhayeri et al [25]. Baltaci and Gokcay used pupil diameter and temperature features in stress detection [26], while Choi used HRV, respiration, GSR, EMG, acceleration, and geographical location [27]. New noncontact methods have also been developed recently to measure stress states. Some of them are hyperspectral imaging technique [28], human voice [29, 30], pupil diameter [31], visible spectrum camera [32], or using stereo thermal and visible sensors [33]. However, observing several markers for identifying stress requires an increasing number of input sensors which in turn increases the overall price and lowers applicability. Prices for heart rate meters range from $70 to $500 USD; GSR devices range from $100 to $500 USD, while EMG devices have price ranges from $450 USD up to $1750 USD. Systems combining multiple sensors are priced much higher. For such systems prices fall between $550 USD and $5700 USD, which already can be considered excessive for a mass telemedical lifestyle counseling application. Therefore, in an ambient assisted living (AAL) system, the number of input sensors should be kept minimal. In the rest of the paper, we focus on the simplest and most researched sensor input, that is, the electrical activity of the heart. As for the reliability of HRV sensors, there are still surprisingly few reviews reported in the literature to date on the validation of the information content of low cost sensors compared to a clinically accepted “gold standard” device. Some devices that were tested for validity are the SenseWear HR Armband [34], the Smart Health Watch [35], the Actiheart [36, 37], the Equivital LifeMonitor [38], and the PulseOn [39]; and also the Bioharness multivariable monitoring device from Zephyr has been tested for validity [40, 41] and reliability [41, 42]. In all cases, a gold standard
device was used simultaneously with the device under test as a method for validating data. However, the validated devices above are high-end devices with a considerable price which present an obstacle for the penetration of telemedicine. For example, the Bioharness device has a price around $550 USD, whereas the price of low cost heart rate meters varies from $70 USD to $100 USD. The lack of reliability tests of low cost devices was our motivation for our device validation study. For automated stress detection, several methods have been published which use only HRV. In 2008, Kim et al. collected HRV data from sixty-eight subjects [43]. HRV data were collected during three different time periods. High stress decreased HRV features. A maximum classification accuracy of 66.1% was achieved. Melillo et al. in 2011 used nonlinear features of HRV for real-life stress detection [44]. HRV data were collected two times, during university examination and after holidays, on 42 students. Most of HRV features significantly decreased during stress period. Stress detection with classification accuracy of 90% was reported using two Poincar´e plot features and Approximate Entropy. One year later, using the same data, they designed a classification tree for automatic stress detection based on LF and pNN50 HRV features with sensitivity of 83.33% [45]. In 2013, Karthikeyan et al. created stress detection classifiers from ECG signal and HRV features [46]. Vanitha and Suresh used a hierarchical classifier to classify stress into four levels with a classification efficiency of 92% [47] in 2014. In 2015 Munla et al. used an SVM-RBF classifier to predict driver stress with an accuracy of 83% [48].
2. Methods The main goal of this study is the development of a reliable, robust, low price stress detection method suitable for mobile health applications. The study included two distinct phases. In the first phase (device validation study) we tested the reliability of a low cost telemedical heart rate sensor against an accepted medical device. In the second phase we performed and evaluated a clinical study, using the validated telemedical sensor. 2.1. Device Validation Study 2.1.1. Sensor Selection and Measurement Protocol. Among many low cost devices, we have chosen and analyzed CardioSport TP3 Heart Rate Transmitter device, a simple commercial chest belt, as a source of heart rate data, because this is one of the few devices that can measure both heart rate and millisecond accurate RR time interval data. Since this device does not have its own memory for storing data, we used a Nexus 7 tablet with Android version 4.4.2 to connect to the device with the bluetooth 4.0 protocol and store the measured data on the tablet. The reference “gold standard” device was a Schiller MT-101/MT-200 Holter device which was designed for clinical use (see Figure 1). Five healthy male volunteers used the two devices simultaneously during a 24-hour long period in order to make the measurements (see Figure 1). For chest belt sensors, the temporary detachment or dislocation of the sensor during
Journal of Healthcare Engineering
3
(a)
(b)
Figure 1: The Schiller MT-101/MT-200 device (a) and the CardioSport TP3 Heart Rate Transmitter device (b).
physical activity or sleep is a common source of errors according to our experiences. Though this problem could be mitigated by using tapes for fixing the device firmly to the body, we felt that such discomfort would not be tolerated in a real AAL situation, so we did not use tapes and used only the daytime 12 hours of the overall signal for analysis. After the monitoring period, we collected the devices and stored the measured data in a unified database. The protocol was reviewed and approved by the institutional ethics review board in January 2014. The volunteers expressed their informed consent to participate and expressed that they understand the goals of the study before the experiment. The comparison of the measured data was a hard task due to the different designs of the gold standard and the telemedical device. However, we wanted to compare signals directly in the time domain and also to develop a data cleaning algorithm for the removal of the noisy parts of the CardioSport device measurements, without using the gold standard data. As the chest belt is not firmly attached to the body, even a slight movement of the device could sometimes cause signal loss (especially during sleep). Therefore, we created a software module for synchronization and data cleaning before any further analysis. Data cleaning meant to remove obviously bad data (artifacts) and to keep only “good” data segments of sufficient length, because, as a rule of thumb, both HRV and Poincar´e plot computation require data chunks of at least 5 minutes. Even though the data cleaning algorithm removes lots of data from the original signal, such a procedure poses no great obstacles for further calculations since we still have enough “good” data during daytime. 2.1.2. Synchronization Procedure. Since the time stamps of the measured records can shift due to device buffering, we used a simple procedure to synchronize the data measured by the CardioSport device with those measured by the gold standard device in order to facilitate their comparison. The algorithm
uses a sliding window that passes from the beginning of the chest belt signal to the end and calculates the absolute error between the two signals. When sliding is over, the location of the sliding window with the minimum absolute error is considered as the point where the two signals should be synchronized. This applies only if the correlation of the data in the sliding window and the same amount of data from the gold standard are higher than a minimum set by the user. If these conditions are met, the algorithm copies data from the sliding window into a newly generated third signal which represents the chest belt signal fully synchronized with the gold standard signal. If conditions are not met, the third signal is filled with zeros. At the end, the algorithm extracts all the highly correlated segments from the third signal skipping zero values. Also, a file with all the merged segments is generated for general analysis. The algorithm uses the following 5 main parameters, with their values determined empirically in parentheses: (i) Window size: how much data is copied from the signal into the sliding window (default: 200). (ii) Window shift step: number of samples by which we shift the sliding window in each iteration (default: 50). (iii) Absolute error window: how much data will be used to calculate the minimum absolute error (default: 200). (iv) Maximum error distance: number of samples by which we shift the absolute error window in order to find the minimum absolute error (default: 1000). (v) Minimum correlation: minimum correlation, expressed as a percentage, required for the two signals to consider data in the chest belt signal accurate (default: 97%). 2.1.3. Statistical Analysis and Data Processing. We performed time- and frequency-domain analysis and computed the correlation and mean absolute percentage error of the two
4 measurements. We also compared the slope of the scatter plot diagrams of the two measurements. The time- and frequency-domain analysis for HRV was performed in Kubios HRV analysis software, while the rest of the analysis was performed in Microsoft Excel. We developed a simple data cleaning algorithm to be used in a real telemedical scenario, for automatically finding good parts of the signal, even without gold standard data. This means finding gaps and abnormal values and skipping them. First, we compare the timestamp of each data item with the timestamp of the previous one. If the difference between the timestamps is bigger than 3 seconds, we mark this data as a gap. Three seconds is used for gap detection because the chest belt has a buffering system that can tolerate short detachments of the device from the body. If more than 3 seconds is used, some data could be missing which could cause errors in further data analysis. In the second step we identify abnormal values in the signal. It is important to emphasize that we do not modify the data in any way as this could potentially result in false results in the subsequent analysis. Instead, abnormal values are treated the same way as gaps. The abnormal values are identified by observing the mean value of 20 contiguous samples (10 previous and 10 following ones). If this mean value differs from the value of the current sample by more than 300, we consider it invalid and mark as gap/error in the signal. Finally, we extract the good segments from the signal with a length of more than 5 minutes. 2.2. Clinical Study 2.2.1. Measurement Protocol. 46 healthy volunteers, mostly university and high school students (27 men and 19 women; average age: 24.6 years), participated in the experiment. The experiment was divided into two parts with a duration of 10 minutes each, so the whole procedure lasted for 20 minutes. In the first part, the participants were asked to try to relax in upright sitting position while listening to relaxation music. The second part of the experiment was a mental task designed to serve as a source of mental stress. We used the Stroop color test smartphone game [49] which is commonly applied to induce mental stress in similar studies. In this game, the user must connect colors to labels at an ever increasing pace. Since controlled breathing and posture have been reported before influencing HRV features, we asked the participants not to control their breathing and to sit still in the same position during the whole experiment. This was also necessary to prevent the detachment of the chest belt from the body. RR intervals were recorded using the CardioSport TP3 Heart Rate Transmitter. The participant was asked about her/his subjective stress levels on a relative scale three times, that is, before the experiment, after the relaxation part, and after the game playing part. The answers along with her/his age and gender were recorded in a simple questionnaire (see Appendices A-B). The reason for such questions was that, though less expected, game playing may be more relaxing for some people than music, and only if we actually succeeded in raising the stress level in the second part compared to the first part, can we expect any algorithm or method to detect the
Journal of Healthcare Engineering stress. After the recording, the device was unmounted from the participant. The protocol was reviewed and approved by the institutional ethics review board in January 2014. The volunteers expressed their informed consent to participate and expressed that they understand the goals of the study before the experiment. After the experiment all data was stored in a unified database, and the data cleaning algorithm described in Section 2.1 was run on both 10-minute parts of each record. Those participants whose records contained no “clean” segments of at least 5 minutes in both parts were excluded from the further analysis. Similarly, we excluded those who— despite our efforts for provoking stress in the experiment— reported no increase of stress level due to game playing. 2.2.2. Statistical Analysis. We used the Kubios software package for getting HRV features and later we analyzed and compared data using the MedCalc software and Microsoft Excel. Wilcoxon paired-samples test was used as a tool for determining significant changes between the two parts of the experiment for the measured values of the HRV features and a 𝑝 value of < 0.05 was considered as significant. Correlation, percentage differences, average percentage differences, and minimum percentage differences were also calculated for all the observed HRV features. 2.2.3. Stress Detection Algorithm. We developed a simple algorithm to detect stress that uses only time-domain HRV features. The reason for excluding frequency-domain features is that they require much more computing power to calculate than time-domain features, an argument that we should consider in a solution designed for mobile devices. We used a combination of the mean HR, pNN50, and RMSSD features to identify stress. A sliding window over the HR signal was divided into four equal parts. We tested various lengths of the sliding window and the shortest width of the window that achieved good result was 560 RR intervals with shift of 20 RR intervals in each step. We used brute force technique to find best threshold values for each HRV feature. As a result, stress is detected by the algorithm if the mean heart rate in the fourth part compared to the first part increases by more than 5%, and RMSSD and pNN50 values decrease by more than 9% in the fourth part compared to the third part. We must emphasize here that this algorithm does not detect rest state. So for the sake of calculating accuracy, specificity, and sensitivity, rest state is considered if stress was not detected. The state of the subject after stressor is also not recognized. Therefore, instead of detecting the subject’s physiological state of stress, the purpose of our algorithm is to detect those stressful events which have negative impact on the subject’s current state but which may or may not lead the subject into a stressful state. A series of stressful events instead of a single major event can also gradually put the subject into a stressful condition. In a binary classification model this could lead to the false conclusion that only the last event was the one which caused stress, while all the previous events are not taken into consideration and remain hidden.
Journal of Healthcare Engineering
5
Start
Calculate mean heart rate of the first and the fourth part (meanHR_1, meanHR_4)
meanHR_4 > 1.05∗ meanHR_1?
Yes
Calculate RMSSD value of the third and fourth part (RMSSD_3, RMSSD_4)
RMSSD_3 > 1.09∗ RMSSD_4?
Calculate pNN50 value of the third and fourth part (pNN50_3, pNN50_4)
Yes
pNN50_3 > 1.09 ∗ pNN50_4?
Yes
Stress
Figure 2: Flow chart of stress detection algorithm.
We note that though it is true that the HRV features may be due to other factors such as depression and mood, we postulate that these factors do not change during the experiment. In contrast, the proposed method uses a welldefined change in the HRV features to detect the beginning of the (induced) stressful state; therefore we expect no false positive stress detections due to such factors. Figure 2 shows a summary flow chart of the proposed stress detection algorithm which implements the above procedure. In order to test the power of this algorithm, we compared its performance to two state-of-the-art algorithms from the same author, Melillo et al. [31, 32], on the same dataset. 2.2.4. Performance Comparison to a Linear HRV Algorithm. The algorithm described in [45] uses the pNN50 feature from the time domain and the LF feature from the frequency domain to create a simple classification tree. Stress is detected if LF < 899.58 and pNN50 > 0.9873 or if LF > 277.28 and pNN50 < 0.9873. Restful state is detected if LF > 899.59 or if LF > 277.28 and pNN50 < 0.9783. For extracting the LF and pNN50 features, we used the same software as the authors of this algorithm, Kubios. Our experiment consisted of two 10 minutes long periods so we extracted two 5 minutes long segments from relaxation part and two 5-minute segments from game playing part. If stress was detected in any one of them, we marked the whole 10minute part as STRESS. If both parts were detected as REST, then whole 10-minute part was marked as REST. 2.2.5. Performance Comparison to a Nonlinear HRV Algorithm. A stress detection method based on nonlinear analysis [44] was the next algorithm we used. This algorithm uses three nonlinear features: Poincar´e plot SD1, Poincar´e plot SD2, and Approximate Entropy (En). According to the method, stress is found if 10.64 + 203.99 ⋅ SD1 − 108.74 ⋅ SD2 − 8.26 ⋅ En (0.2) > 0.
(1)
Table 1: Signal durations after the synchronization process. Subject number Duration (hh:mm:ss)
#1
#2
#3
#4
#5
2:06:18
10:53:28
8:45:40
10:30:17
7:46:56
To reconstruct this algorithm we used Microsoft Visual C# to calculate Approximate Entropy based on formula described by authors. A sliding window was used to scan the whole relaxation part as well as the game playing part. If stress was found in any step, we marked the whole 10-minute period as STRESS and, similarly, if rest was detected on all steps of whole part, we marked that part as REST. In order to compare the performances of the three methods, we computed the accuracy, specificity, and sensitivity for each of them. For this, we registered a true positive result if the method marked the game playing part as STRESS, a true negative result if the relax music part was marked as REST, a false positive result if the relax music part was marked as STRESS, and a false negative result if the game playing part was marked as REST.
3. Results 3.1. Device Validation Study: Comparison with the Gold Standard. After running the synchronization process, we got segments of highly correlated data. Figure 1 shows how the lengths of signal segments are distributed. We can see that most segments are 3–18 minutes long. The longest segment that is highly correlated with the gold standard data is 110 minutes long. The default parameter settings minimize the number of overly short (<5 min) segments. Most of the bad segments (Figure 3) are shorter than one minute, and only one bad segment was 60 minutes long. The synchronization procedure resulted in highly (>97%) correlated synchronized data segments with various durations. Table 1 shows the overall duration of signals. Subject #1
Journal of Healthcare Engineering 14
90
12
80 Number of segments
Number of segments
6
10 8 6 4
70 60 50 40 30 20
2 0
10 0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
1
2
3
4
5
(Minutes) Minutes
6
7 8 9 10 11 12 13 14 15 (Minutes)
Minutes (a)
(b)
Figure 3: Distribution of highly (a) and low correlated (b) segment lengths for all subjects after synchronization procedure. Table 2: Time-domain results after the synchronization process. Subject number #1 #2 #3 #4 #5 Average
Schiller 738.27 704.04 907.63 854.53 937.01 850.80
Mean RR (ms) CardioSport 755.47 720.42 928.88 874.50 958.97 870.69
Error 2.28% 2.27% 2.29% 2.28% 2.29% 2.28%
had the lowest usable time with only 2 hours and 6 minutes. The most probable reason for such a low time is the chest hair which reduced the contact between electrodes and the skin. For this reason, this subject was excluded from calculation of average results. Table 2 shows results in time domain for the Schiller and the CardioSport devices after using our algorithm for the synchronization of signals. The time-domain analysis shows pretty close values for both mean RR values and standard deviation. The formula used for computing the standard deviation of RR intervals is as follows: STD RR = √
2 1 𝑁 ∑ (RR𝑗 − RR) . 𝑁 − 1 𝑗=1
(2)
Average mean RR values for Schiller and CardioSport devices are 850.80 and 870.69, respectively. Average STD RR for the Schiller device is 108.42 and it is 110.93 for the CardioSport device. The frequency-domain analysis is presented in Table 3. The absolute power was compared for very low frequency (VLF: 0–0.04 Hz), low frequency (LF: 0.04–0.5 Hz), and high frequency (HF: 0.15–0.4 Hz) and ratio between low frequency and high frequency (LF/HF). The results show no significant difference between Schiller and CardioSport device values.
Schiller 123.34 91.35 90.40 144.74 107.18 108.42
STD RR (ms) CardioSport 125.09 93.47 92.83 148.00 109.41 110.93
Error 1.40% 2.27% 2.62% 2.20% 2.04% 2.11%
The average mean absolute percentage error (MAPE) between the two signals is 2.32% with a high average correlation of 99.67%. 3.2. Device Validation Study: Data Cleaning Method. We run the data cleaning algorithm described in Section 2 on the data recorded by the chest belt. The duration of the resulting signal is shown in Table 4. Similar to the synchronization process, we got a very short duration for one subject and we excluded this subject from further analysis. It is important to note that, due to the noise on Schiller device records, we had to remove noisy parts from the “gold standard” signal as well. Therefore, even though the signal was recorded for 12 hours continuously, the overall duration is much less. The calculation shows that, in the worst scenario, only 45% of the signal can be used for analysis using this data cleaning method. However, in the best scenario, this number reaches 95%. This leads to a conclusion that results are quite subject dependent. Table 5 shows the results of data analysis in the time domain after removing bad parts with the data cleaning algorithm. We can see that the mean RR intervals for the Schiller and the CardioSport devices are 851.14 and 871.23 and the standard deviations are 104.61 and 106.35, respectively. In general, the CardioSport device has slightly greater values but they are very close.
Journal of Healthcare Engineering
7
Table 3: Frequency-domain analysis after the synchronization process. Subject number VLF 7937.6 5431.5 4251.2 12682 6139.8
#1 #2 #3 #4 #5
Schiller Absolute power (ms2 ) LF HF 3086 1578 626.6 245 1927 494.4 1790 636.5 1212 476.7
LF/HF 1.956 2.557 3.898 2.813 2.542
VLF 8444 5723 4543 13514 6465
Table 4: Signal durations after the data cleaning process. Subject number Duration (hh:mm:ss)
#1
#2
#3
#4
#5
1:28:10
11:20:03
6:15:38
9:27:07
4:29:44
Schiller versus CardioSport comparison—subject #5 1600 y = 1.0124x + 10.799 R2 = 0.9875
1400
CardioSport
1200 1000 800 600 400 200 0 0
200
400
600
800
1000
1200
1400
Schiller
Figure 4: Comparison of CardioSport and Schiller device after data cleaning.
The minimum, maximum, and average percentage errors on the whole signal were calculated using a 5 minutes long sliding window with one-minute shift step (Table 6). Only one subject had a very high maximum error value of 33.86%. By visual examination, we determined that the cause of this high error percentage was in fact the presence of artifacts in the “gold standard” Schiller device measurements. Despite that, average error values are at a very low level of 2.20%. Figure 4 demonstrates the typical relationship between the CardioSport and Schiller measurements using a scatter plot for subject #5. All slope values are close to 1. The lowest slope value is 0.9757, while the highest value is 1.0184. The average mean absolute percentage error (MAPE) between the two signals was 2.62% with a high average correlation of 98.76%.
CardioSport Absolute power (ms2 ) LF HF LF/HF 3224 1330 2.4235 659.3 250.9 2.6281 2055 538.8 3.8146 1869 621.5 3.0077 1274 481.4 2.6459
% VLF 6.00 5.09 6.42 6.16 5.03
% LF 4.28 4.96 6.23 4.23 4.87
Error % HF 18.65 2.35 8.24 2.41 0.98
% LF/HF 19.29 2.71 2.19 6.47 3.93
3.3. Clinical Study. Five subjects were excluded from further analysis because they reported a decrease (instead of the expected increase) of their stress level while playing the Stroop game. As an explanation, some participants reported that playing the game was much more joyful than relaxation music. Others reported that the game kept their mind focused and that the relaxation music brought them back to their problems and duties of the day. Some also reported anxiety about the experiment itself which vanished while playing. After removing these records, we run the data cleaning algorithm which identified 10 noisy records, probably due to too much movement. These were also excluded, so the active dataset decreased to 31 subjects’ records (20 men and 11 women; average age = 24.7 years). Table 7 shows 𝑝 values of the Wilcoxon paired-samples test, for the relax versus stress parts, for the relevant HRV features (𝑛 = 31). We found a statistically significant difference for the following time-domain features: mean RR (𝑝 = 0.0001), mean HR (𝑝 = 0.0001), pNN50 (𝑝 = 0.0103), NN50 (𝑝 = 0.0128), RMSSD (𝑝 = 0.0255), and HRV triangular index (𝑝 = 0.0456). In frequency domain, two features showed statistically significant difference: HF (ms2 ) with 𝑝 = 0.0054 and LF (ms2 ) with 𝑝 = 0.0128. The VLF (%) feature was also close but not significantly different (𝑝 = 0.0745). In nonlinear analysis, the SD1 feature showed a statistically significant difference (𝑝 = 0.0268). The average percentage differences and the minimum percentage differences are shown in Table 8. Table 9 shows the correlations between the important features during the relaxation part of experiment. We can see very high positive correlation (higher than 0.9) between the following features: NN50 and RMSSD (0.94818), pNN50 and RMSSD (0.935664), and pNN50 and NN50 (0.98966). Poincar´e plot SD1 feature was highly correlated with HF (ms2) feature. Only one very high negative correlation was found between features mean RR and mean HR (−0.99452). Figure 5 shows, as an example, the values of an observed feature (mean HR) for the relaxation period and the game playing period, respectively. 3.4. Clinical Study: Stress Detection Performance Compared to Other Methods. The accuracy, sensitivity, and specificity values for correctly detecting stress are shown in Table 10 for
8
Journal of Healthcare Engineering Table 5: Time-domain analysis after the data cleaning process.
Subject number
Schiller 707.80 700.40 899.20 846.46 958.49 851.14
#1 #2 #3 #4 #5 Average
Mean RR (ms) CardioSport 724.03 716.70 920.97 866.25 981.00 871.23
Table 6: Minimum, maximum, and average percentage error. Subject number
Minimum error
Maximum error
Average error
#1 #2 #3 #4 #5 Average
0.08% 0.01% 0.04% 0.13% 0.07% 0.06%
3.50% 7.71% 33.86% 6.72% 5.11% 13.35%
1.50% 2.12% 3.22% 1.92% 2.22% 2.37%
Table 7: Statistical significance of the observed features ordered by 𝑝 value. Feature Mean RR Mean HR HF (ms2 ) pNN50 NN50 LF (ms2 ) RMSSD Poincar´e plot, SD1 HRV triangular index VLF (%) STD RR HF (%) Poincar´e plot, SD2 LF/HF VLF (ms2 ) TINN Power (n.u.)-HF Power (n.u.)-LF STD HR
𝑝 value 0.0001 0.0001 0.0054 0.0103 0.0128 0.0128 0.0255 0.0268 0.0456 0.0745 0.1583 0.1583 0.2725 0.4565 0.4565 0.5967 0.7390 0.7539 0.9687
our algorithm, the linear algorithm proposed by Melillo et al., and the nonlinear algorithm proposed by the same authors.
4. Discussion Stress is a very complex subject and measuring stress is not an easy task. There are many markers that could be used, many algorithms that could be applied, and many forms of
Error 2.24% 2.27% 2.36% 2.28% 2.29% 2.29%
Schiller 136.04 91.33 99.67 139.26 88.16 104.61
STD RR (ms) CardioSport 138.63 93.24 99.77 142.33 90.08 106.35
Error 1.87% 2.05% 0.10% 2.16% 2.13% 1.66%
Table 8: Average percentage difference and minimum percentage difference for the features computed from the HR signal. Feature
Average percentage difference
Minimum percentage difference
6.88 27.86 72.76
0.94 3.98 3.88
Mean HR RMSSD pNN50
120 100 80 60 40 20 0 1
3
5
7
9 11 13 15 17 19 21 23 25 27 29 31
Mean HR-relaxation Mean HR-stress
Figure 5: Mean HR feature for all subjects during relaxation and while playing game.
stress which could be observed. Heart rate variability, being simple and noninvasive, has recently become one of the most popular methods for detecting stress. Still, this is not an easy task, since HRV is not a single value; rather, it consists of many features that can be observed in time domain and frequency domain or using nonlinear analysis. The literature generally reports that, under mental stress, the mean RR, pNN50, STD RR, and RMSSD features decrease, while the mean HR and LF features increase significantly. However, significant differences for the same features and sometimes even opposite results (e.g., LF feature) are also reported. One probable cause for this inconsistency in literature could be the fact that stress is not the only condition that influences changes in HRV. Physical activity, body posture, breathing, age, gender, and illnesses all have a great influence on HRV. In this paper, we analyzed various HRV features in order to find those that change significantly under mental stress and proposed a simple stress detection algorithm.
Journal of Healthcare Engineering
9
Table 9: Correlation of observed features during relaxation part.
Mean RR Mean HR RMSSD NN50 pNN50 HRV t.i. LF (ms2 ) HF (ms2 ) P.P., SD1
Mean RR 1.00 −0.99 0.29 0.38 0.48 0.25 0.08 0.20 0.12
Mean HR
RMSSD
NN50
pNN50
1.00 −0.28 −0.37 −0.47 −0.25 −0.09 −0.20 −0.13
1.00 0.95 0.94 0.78 0.15 −0.09 −0.01
1.00 0.99 0.77 0.17 −0.03 0.01
1.00 0.75 0.16 0.00 0.03
Table 10: Performance comparison of the three stress detection methods. Feature Accuracy Sensitivity Specificity
Melillo linear 61.29% 61.29% 61.29%
Melillo nonlinear 50.00% 29.03% 70.97%
Our method 74.60% 75.00% 74.19%
In the device validation study, we tested the reliability of a low cost heart rate meter. The CardioSport TP3 heart rate meter device was used simultaneously with a professional ECG recorder (Holter) device. We compared the results using standard deviation, correlation, and scatter plot diagram with slope of the regression line which are commonly used in literature [36, 40, 50]. However, before we analyzed the results, we used a simple data cleaning algorithm to eliminate noisy parts without using any data correction. The data cleaning process reduces the overall duration of the signal but it increases its quality. After data cleaning, all results of the CardioSport device were very close to the Schiller device with an average correlation of 98.73%. The downside of the data cleaning algorithm is that it will delete sections of the signal even with the slightest detachment. Hopefully, with advance in wearable sensors, new forms of heart rate monitors like rings or bracelets with firmer attachment to body will be available. In the next step, we demonstrated how a simple mental stressor can influence HRV features significantly. Our findings are not very different from previous research, showing that HRV can indeed be used as an indicator of mental stress. We found that, under the influence of mental stress, mean HR increased, while mean RR, pNN50, RMSSD, and HRV triangular index decreased. Contrary to the results from literature, we did not find a statistically significant difference in STD RR feature (𝑝 = 0.1583). This could be explained by the fact that we analyzed only 10 minutes in each part of the experiment, while STD RR feature describes long-term variability. A limitation of our study is that we only analyzed the influence of mental stress. Physical or emotional stress could influence observed features in a completely different way.
HRV t.i.
LF (ms2 )
HF (ms2 )
P.P., SD1
1.00 0.28 −0.21 −0.15
1.00 0.33 0.51
1.00 0.89
1.00
For some subjects, the experiment failed to provoke mental stress which is of course a shortcoming of the experiment setup; however, it would be very hard to design a method that is successful in all cases. The Stroop test was chosen because it is easy to implement and is generally accepted in the literature for such purposes. Since it increases the speed of the game proportionally with the user’s results, it should increase the stress level regardless of the subjects’ cognitive level. We believe that other factors, such as the subject’s prior experience and motivation for playing computer games, are harder to control. As a conclusion of this study we created a robust stress detection algorithm. Unlike other stress detection algorithms which use several stress markers [51, 52], we used only HRV features for stress detection but with relatively high stress identification ratio. We were able to get an accuracy rate of 74.60%, somewhat below the 85% reported by the algorithm in [53]; however, the latter uses the full electrocardiogram (ECG) measurement compared to using only RR intervals in our case. If we compare our algorithm with other algorithms for stress detection using only HR and HRV parameters, we can say that we achieved higher identification rate than the algorithm in [43] and a worse result than [48] or [45] and around 15% worse compared to [44, 47] but we used only time-domain features for stress identification instead of frequency-domain features or using nonlinear analysis which are much more expensive to calculate. Since performance of particular algorithm depends on multiple parameters like type of stressor, number of subjects, methods used, and so forth, we compared the performance of our algorithm with two state-of-the-art algorithms. Although our algorithm showed lower declared accuracy, the comparison of its performance on the same dataset showed much better results than the algorithm that uses nonlinear HRV features [44] and slightly better performance than the algorithm that uses linear HRV [45] features for stress detection. We think that the poor performance of the two tested methods could partly be due to the fact that our stressor, the Stroop game, was not as strong as the university exam stressor used for the development of the Melillo algorithms or driving [48]. Such results show that the same algorithm can give very different results on different dataset meaning that comparing
10
Journal of Healthcare Engineering
strictly by accuracy values is not good indicator if comparison is not performed on the same dataset. A weakness of the proposed algorithm is that it only detects events that provoke stress for a particular subject. It cannot classify the current state of a subject in given moment. However, this was not the intention of our algorithm from the beginning and we propose using combination of our algorithm with typical classification algorithms for achieving greater insight into stressful events and a subject’s current state. A strength of our algorithm is that an event or a series of stressful events could be detected before entering into a stressful state defined by classification algorithms. The user could be informed by sound, vibration, or other kinds of alert when stressful event happens, leading to greater awareness about daily stressors. Also, this algorithm is very simple and easy to implement in mobile environments due to the fact that all HRV features are chosen only from time domain.
5. Conclusions From this research we can conclude that even a simple low cost heart rate monitor device can detect features that change significantly under the influence of mental stress. Using these results we created a simple stress detection algorithm that is being integrated in the Lavinia lifestyle counseling mobile application [54] for further testing and refinement in real-life stress situations. If stress detection proves to be reliable for larger samples, it will be used in the blood glucose prediction models developed for diabetics.
Appendix A. The Instructions for the Clinical Study Experiment: Measuring Physiological Effects of Heart Rate Parameters during Various Situations Instructions to Participants. During this experiment we will measure physiological parameters of various situations. First, we will measure parameters during listening to relaxation music and then we will measure the same parameters during playing simple Android based game using mobile phone. You will need to fill data in the questionnaires three times: before experiment starts, after the relaxation music stops, and at the end of experiment. If you feel any discomfort during this experiment, you can stop at any time: Personal Data Name or code: — Age: — Gender: M/F How to Play. In this game screen is divided in four parts. In each part word is shown with different color. Words are color names like red, blue, purple, black, green,. . .. You need to touch the word which is colored in the color of the word. For example, if word red is colored with black, then you skip
it, but if the word red is colored in red color, then you need to press that word. If you make a mistake, game ends. You need to get as high score as possible. For every choice, you have only a limited amount of time. If time runs out, game is over. You can play as many games as you can during 10 minutes: Example
Red Green Correct
Blue Black Incorrect
Word red is colored red so this is correct choice. Word green is colored with black color so this is incorrect choice. Word blue is colored with orange color so this is incorrect choice. Word black is colored with blue color so this is incorrect choice.
B. Experiment Data and Questionnaire Experiment Data Date: — Protocol Give person the paper with experiment description. Let him/her read whole description (including “how to play” section). Ask if they understand the experiment. Tell the person that in this experiment they will first listen one relaxation song and then they will play simple game. Show the person how to play game shortly if necessary. Explain to person that during whole experiment he/she will need to wear chest belt for collecting heart rate data. Ask the person to put chest belt. Ensure privacy for person to put chest belt on. Time (hh:mm): — Establish bluetooth connection between chest belt and Android device. Write down time when this occurs. Time (hh:mm): — Ask the person about current perceived stress level from 1 to 10 where 1 is no stress and 10 is maximum stress level and write down perceived stress level Time (hh:mm): — Perceived stress level: — Tell the person to relax while listening relaxation music. Start playing relaxation music and write the exact start time (when you hit the play button) Time (hh:mm): —
Journal of Healthcare Engineering When music stops, write down the time of the event. Ask the person about perceived stress level after listening the song from 1 to 10 where 1 is no stress and 10 is maximum stress level. Write down perceived stress level Time (hh:mm): — Perceived stress level: — Ask person to play Android based game for the next 10 minutes. Write the start time and set countdown timer to at least 10 minutes. Time (hh:mm): — At the end of 10 minutes, tell person to stop playing and write down the time. Ask person about perceived stress level after playing game from 1 to 10 where 1 is no stress and 10 is maximum stress level. Write down perceived stress level Time (hh:mm): — Perceived stress level: —
Abbreviations AAL: BP: ECG: EMG: GSR: HF: HR: HRV: LF: MAPE: NN50:
Ambient assisted living Blood pressure Electrocardiograph Electromyogram Galvanic skin response High frequency Heart rate Heart rate variability Low frequency Average mean absolute percentage The number of pairs of successive beat-to-beat or NN intervals that differ by more than 50 ms pNN50: The proportion of NN50 divided by total number of NNs RMSSD: The square root of the mean of the squares of the successive differences between adjacent intervals RR (interval): R wave to R wave interval VLF: Very low frequency.
Competing Interests The authors declare that there are no competing interests regarding the publication of this paper.
References [1] A. Mitra, “Diabetes and stress: a review,” Ethnomedicine, vol. 2, no. 2, pp. 131–135, 2008. [2] A. Steptoe and M. Kivim¨aki, “Stress and cardiovascular disease: an update on current knowledge,” Annual Review of Public Health, vol. 34, no. 1, pp. 337–354, 2013.
11 [3] A. M. Heraclides, T. Chandola, D. R. Witte, and E. J. Brunner, “Work stress, obesity and the risk of type 2 diabetes: genderspecific bidirectional effect in the whitehall II study,” Obesity, vol. 20, no. 2, pp. 428–433, 2012. [4] Y.-M. Oh, Y. S. Kim, S. H. Yoo, S. K. Kim, and D. S. Kim, “Association between stress and asthma symptoms: a population-based study,” Respirology, vol. 9, no. 3, pp. 363–368, 2004. [5] S. Patel, H. Park, P. Bonato, L. Chan, and M. Rodgers, “A review of wearable sensors and systems with application in rehabilitation,” Journal of NeuroEngineering and Rehabilitation, vol. 9, no. 1, article 21, 2012. [6] A. Pantelopoulos and N. G. Bourbakis, “A survey on wearable sensor-based systems for health monitoring and prognosis,” IEEE Transactions on Systems, Man and Cybernetics Part C: Applications and Reviews, vol. 40, no. 1, pp. 1–12, 2010. [7] A. Pantelopoulos and N. Bourbakis, “A survey on wearable biosensor systems for health monitoring,” in Proceedings of the 30th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBS ’08), pp. 4887–4890, Vancouver, Canada, August 2008. [8] S. Ortmann, P. Langendorfer, and C. S. Lanyi, “Telemedical assistance for ambulant rehabilitation of stroke patients,” Brain Injury, vol. 26, Meeting Abstract 0650, no. 4-5, pp. 644–645, 2012. [9] Y. Zhang, H. Liu, X. Su, P. Jiang, and D. Wei, “Remote mobile health monitoring system based on smart phone and browser/server structure,” Journal of Healthcare Engineering, vol. 6, no. 4, pp. 717–738, 2015. [10] Y. Shi, N. Ruiz, R. Taib, E. Choi, and F. Chen, “Galvanic skin response (GSR) as an index of cognitive load,” in Proceedings of the Extended Abstracts on Human Factors in Computing Systems (CHI EA ’07), San Jose, Calif, USA, 2007. [11] U. Lundberg, R. Kadefors, B. Melin et al., “Psychophysiological stress and emg activity of the trapezius muscle,” International Journal of Behavioral Medicine, vol. 1, no. 4, pp. 354–370, 1994. [12] J. Wijsman, B. Grundlehner, J. Penders, and H. Hermens, “Trapezius muscle EMG as predictor of mental stress,” in Proceedings of the 1st Wireless Health Conference (WH ’10), pp. 155–163, ACM, October 2010. [13] T. Yamakoshi, K. Yamakoshi, S. Tanaka et al., “Feasibility study on driver’s stress detection from differential skin temperature measurement,” in Proceedings of the 30th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBS ’08), pp. 1076–1079, Vancouver, Canada, August 2008. [14] J. Taelman, S. Vandeput, A. Spaepen, and S. V. Huffel, “Influence of mental stress on heart rate and heart rate variability,” in Proceedings 4th European Conference of the International Federation for Medical and Biological Engineering (IFMBE ’09), pp. 1366–1369, 2009. [15] E. Ceja, V. Osmani, and O. Mayora, “Automatic stress detection in working environments from smartphones’ accelerometer data: a first step,” IEEE Journal of Biomedical and Health Informatics, 2015. [16] S. D. W. Gunawardhane, P. M. D. Silva, D. S. B. Kulathunga, and S. M. K. D. Arunatileka, “Non invasive human stress detection using key stroke dynamics and pattern variations,” in Proceedings of the International Conference on Advances in ICT for Emerging Regions (ICTer ’13), pp. 240–247, Colombo, Sri Lanka, December 2013. [17] A. Marcos-Ramiro, D. Pizarro-Perez, M. Marron-Romera, and D. Gatica-Perez, “Automatic blinking detection towards stress
12
[18]
[19]
[20]
[21]
[22]
[23]
[24]
[25]
[26]
[27]
[28]
[29]
[30]
Journal of Healthcare Engineering discovery,” in Proceedings of the 16th International Conference on Multimodal Interaction (ICMI ’14), pp. 307–310, Istanbul, Turkey, November 2014. A. Fernandes, R. Helawar, R. Lokesh, T. Tari, and A. V. Shahapurkar, “Determination of stress using blood pressure and galvanic skin response,” in Proceedings of the International Conference on Communication and Network Technologies (ICCNT ’14), pp. 165–168, Sivakasi, India, December 2014. F.-T. Sun, C. Kuo, H.-T. Cheng, S. Buthpitiya, P. Collins, and M. Griss, “Activity-aware mental stress detection using physiological sensors,” in Mobile Computing, Applications, and Services, M. Gris and G. Yang, Eds., vol. 76 of Lecture Notes of the Institute for Computer Sciences, Social Informatics and Telecommunications Engineering, pp. 211–230, Springer, New York, NY, USA, 2012. ´ A. De Santos Sierra, C. S. Avila, J. Guerra Casanova, and G. B. Del Pozo, “A stress-detection system based on physiological signals and fuzzy logic,” IEEE Transactions on Industrial Electronics, vol. 58, no. 10, pp. 4857–4865, 2011. G. Rigas, Y. Goletsis, and D. I. Fotiadis, “Real-time driver’s stress event detection,” IEEE Transactions on Intelligent Transportation Systems, vol. 13, no. 1, pp. 221–234, 2012. J. Wijsman, B. Grundlehner, H. Liu, H. Hermens, and J. Penders, “Towards mental stress detection using wearable physiological sensors,” in Proceedings of the Annual International Conference of the IEEE Engineering in Medicine and Biology Society, Boston, Mass, USA, August-September 2011. A. Riera, A. Soria-Frisch, A. Albajes-Eizagirre et al., “Electrophysiological data fusion for stress detection,” Studies in Health Technology and Informatics, vol. 181, pp. 228–232, 2012. M. Singh and A. B. Queyam, “A novel method of stress detection using physiological measurements of automobile drivers,” International Journal of Electronics Engineering, vol. 5, no. 2, pp. 13–20, 2013. F. Mokhayeri, M.-R. Akbarzadeh-T, and S. Toosizadeh, “Mental stress detection using physiological signals based on soft computing techniques,” in Proceedings of the 18th Iranian Conference of Biomedical Engineering (ICBME ’11), pp. 232–237, Tehran, Iran, December 2011. S. Baltaci and D. Gokcay, “Role of pupil dilation and facial temperature features in stress detection,” in Proceedings of the 22nd Signal Processing and Communications Applications Conference (SIU ’14), pp. 1259–1262, Trabzon, Turkey, April 2014. J. Choi, Minimally-invasive wearable sensors and data processing methods for mental stress detection [Ph.D. dissertation], Texas A&M University, 2011, http://hdl.handle.net/1969.1/ ETD-TAMU-2011-12-10674. T. Chen, P. Yuen, M. Richardson, G. Liu, and Z. She, “Detection of psychological stress using a hyperspectral imaging technique,” IEEE Transactions on Affective Computing, vol. 5, no. 4, pp. 391–405, 2014. H. Lu, D. Frauendorfer, M. Rabbi et al., “StressSense: detecting stress in unconstrained acoustic environments using smartphones,” in Proceedings of the 14th International Conference on Ubiquitous Computing (UbiComp ’12), pp. 351–360, ACM, Pittsburgh, Pa, USA, September 2012. A. K. Mishra and O. S. Vaidya, “Attributes detection for allocation of resources with stress detection from speech,” International Journal of Electronics, Electrical and Computational System, vol. 3, no. 4, pp. 61–66, 2014.
[31] P. Ren, A. Barreto, J. Huang, Y. Gao, F. R. Ortega, and M. Adjouadi, “Off-line and on-line stress detection through processing of the pupil diameter signal,” Annals of Biomedical Engineering, vol. 42, no. 1, pp. 162–176, 2014. [32] B. Kaur, S. Moses, M. Luthra, and V. N. Ikonomidou, “Remote stress detection using a visible spectrum camera,” in Independent Component Analyses, Compressive Sampling, Large Data Analyses (LDA), Neural Networks, Biosystems, and Nanoengineering XIII, vol. 9496 of Proceedings of SPIE, May 2015. [33] M. N. Haji Mohd, M. Kashima, K. Sato, and M. Watanabe, “Mental stress recognition based on non-invasive and noncontact measurement from stereo thermal and visible sensors,” International Journal of Affective Engineering, vol. 14, no. 1, pp. 9–17, 2015. [34] M. B. Crawley, Validation of the Sensewear Hr Armband for Measuring Heart Rate and Energy Expenditure, The Pennsylvania State University, 2003. [35] C. M. Lee and M. Gorelick, “Validity of the smarthealth watch to measure heart rate during rest and exercise,” Measurement in Physical Education and Exercise Science, vol. 15, no. 1, pp. 18–25, 2011. [36] J. Kristiansen, M. Korshøj, J. H. Skotte et al., “Comparison of two systems for long-term heart rate variability monitoring in free-living conditions—a pilot study,” BioMedical Engineering OnLine, vol. 10, article 27, 2011. [37] S. Brage, N. Brage, P. W. Franks, U. Ekelund, and N. J. Wareham, “Reliability and validity of the combined heart rate and movement sensor actiheart,” European Journal of Clinical Nutrition, vol. 59, no. 4, pp. 561–570, 2005. [38] Y. Liu, S. H. Zhu, G. H. Wang, F. Ye, and P. Z. Li, “Validity and reliability of multiparameter physiological measurements recorded by the equivital lifemonitor during activities of various intensities,” Journal of Occupational and Environmental Hygiene, vol. 10, no. 2, pp. 78–85, 2013. [39] J. Parak, A. Tarniceriu, P. Renevey, M. Bertschi, R. DelgadoGonzalo, and I. Korhonen, “Evaluation of the beat-to-beat detection accuracy of PulseOn wearable optical heart rate monitor,” in Proceedings of the 37th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC ’15), pp. 8099–8102, Milan, Italy, August 2015. [40] J. A. Johnstone, P. A. Ford, G. Hughes, T. Watson, and A. T. Garrett, “Bioharness6 multivariable monitoring device. Part I: validity,” Journal of Sports Science and Medicine, vol. 11, no. 3, pp. 400–408, 2012. [41] J. A. Johnstone, P. A. Ford, G. Hughes, T. Watson, A. C. S. Mitchell, and A. T. Garrett, “Field based reliability and validity of the bioharness multivariable monitoring device,” Journal of Sports Science and Medicine, vol. 11, no. 4, pp. 643–652, 2012. [42] J. A. Johnstone, P. A. Ford, G. Hughes, T. Watson, and A. T. Garrett, “Bioharness6 multivariable monitoring device: part. II: reliability,” Journal of Sports Science and Medicine, vol. 11, no. 3, pp. 409–417, 2012. [43] D. Kim, Y. Seo, J. Cho, and C.-H. Cho, “Detection of subjects with higher self-reporting stress scores using heart rate variability patterns during the day,” in Proceedings of the 30th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBS ’08), pp. 682–685, August 2008. [44] P. Melillo, M. Bracale, and L. Pecchia, “Nonlinear Heart Rate Variability features for real-life stress detection. Case study: students under stress due to university examination,” BioMedical Engineering Online, vol. 10, no. 1, article 96, 2011.
Journal of Healthcare Engineering [45] P. Melillo, C. Formisano, U. Bracale, and L. Pecchia, “Classification tree for real-life stress detection using linear heart rate variability analysis. Case study: students under stress due to university examination,” in Proceedings of the IFMBE World Congress on Medical Physics and Biomedical Engineering, pp. 477–480, Beijing, China, May 2012. [46] P. Karthikeyan, M. Murugappan, and S. Yaacob, “Analysis of stroop colorword test-based human stress detection using electrocardiography and heart rate variability signals,” Arabian Journal for Science and Engineering, vol. 39, no. 3, pp. 1835–1847, 2014. [47] L. Vanitha and G. R. Suresh, “Hierarchical SVM to detect mental stress in human beings using Heart Rate Variability,” in Proceedings of the 2nd International Conference on Devices, Circuits and Systems (ICDCS ’14), pp. 1–5, Combiatore, India, March 2014. [48] N. Munla, M. Khalil, A. Shahin, and A. Mourad, “Driver stress level detection using HRV analysis,” in Proceedings of the International Conference on Advances in Biomedical Engineering (ICABME ’15), pp. 61–64, Beirut, Lebanon, September 2015. [49] Color Stroop game, February 2016, https://play.google.com/ store/apps/details?id=com.citroon.colortest. [50] A. A. Flatt and M. R. Esco, “Validity of the ithletetm smart phone application for determining ultra-short-term heart rate variability,” Journal of Human Kinetics, vol. 39, no. 1, pp. 85–92, 2013. [51] J. A. Healey and R. W. Picard, “Detecting stress during realworld driving tasks using physiological sensors,” IEEE Transactions on Intelligent Transportation Systems, vol. 6, no. 2, pp. 156–166, 2005. ´ [52] A. De Santos Sierra, C. S. Avila, G. B. Del Pozo, and J. Guerra Casanova, “Stress detection by means of stress physiological template,” in Proceedings of the 3rd World Congress on Nature and Biologically Inspired Computing (NaBIC ’11), pp. 131–136, Salamanca, Spain, October 2011. [53] Listen to Your Heart: Stress Prediction Using Consumer Heart Rate Sensors, February 2016, http://cs229.stanford.edu/ proj2013/LiuUlrich-ListenToYourHeart-StressPredictionUsingConsumerHeartRateSensors.pdf. [54] Lavinia—lifestyle mirror application, http://lavinia.hu/introduction-and-publications/.
13
