Ukusebenziseka kwe-MemTrax kanye Nemodeli Yokufunda Ngomshini Ekuhlukaniseni Ukukhubazeka Kokuqonda Okumaphakathi

Isihloko Sokucwaninga

Ababhali: Bergeron, Michael F. | Landset, Sara | Zhou, Xianbo | Ding, Tao | Khoshgoftaar, Taghi M. | Zhao, Feng | Du, Bo | Chen, Xinjie | Wang, Xuan | Zhong, Lianmei | Liu, Xiaolei| Ashford, J. Wesson

I-DOI: 10.3233/JAD-191340

Ijenali: Ijenali ye Izifo ze-Alzheimer, ivol. I-77, cha. I-4, iphe. 1545-1558, 2020

abstract

Ingemuva:

Ukusabalala kwezigameko kanye nokusabalala kwe Isifo i-Alzheimer kanye nokukhubazeka kwengqondo okumaphakathi (MCI) kwenze ukuthi kubizwe ucingo oluphuthumayo locwaningo ukuze kuqinisekiswe ukutholwa kokuhlolwa kwengqondo nokuhlola kusenesikhathi.

Injongo:

Inhloso yethu eyinhloko yocwaningo bekuwukuthola ukuthi ingabe amamethrikhi okusebenza e-MemTrax akhethiwe kanye nezibalo zabantu ezifanele nezici zephrofayela yezempilo zingasetshenziswa ngempumelelo kumamodeli aqagelayo athuthukiswe ngokufunda komshini ukuze kuhlukaniswe impilo yengqondo (evamile uma iqhathaniswa ne-MCI), njengoba kuzoboniswa Ukuhlolwa Kokuqonda Okuqondwayo kweMontreal (MoCA).

Izindlela:

Senze ucwaningo oluhlukene nge-neurology engama-259, umtholampilo wenkumbulo, kanye neziguli zabantu abadala bemithi yangaphakathi eziqashwe kwababili. izibhedlela eChina. Isiguli ngasinye sanikezwa i-MoCA yolimi lwesiShayina futhi sazilawula ngokwaso ukuqashelwa okuqhubekayo kwe-MemTrax online episodic. inkumbulo test online ngosuku olufanayo. Amamodeli okuhlukanisa okubikezelayo akhiwe kusetshenziswa umshini wokufunda ngokuqinisekiswa okuphindwe izikhathi ezingu-10, futhi ukusebenza kwemodeli kukalwa kusetshenziswa Indawo Engaphansi Kwe-Receiver Operating Characteristic Curve (AUC). Amamodeli akhiwe kusetshenziswa amamethrikhi amabili okusebenza e-MemTrax (amaphesenti alungile, isikhathi sokuphendula), kanye nezici eziyisishiyagalombili ezivamile zokubala kwabantu kanye nezici zomlando womuntu siqu.

Ezenye:

Uma kuqhathaniswa abafundi kuzo zonke izinhlanganisela ezikhethiwe zamaphuzu nama-threshold e-MoCA, i-Naïve Bayes ngokuvamile ibingumfundi owenze kahle kakhulu ngokusebenza kwezigaba kuka-0.9093. Ngaphezu kwalokho, phakathi kwabafundi abathathu abaphezulu, ukusebenza kwezigaba okusekelwe ku-MemTrax kukonke bekuphakeme kusetshenziswa izici ezine ezikleliswe phezulu kuphela (0.9119) uma kuqhathaniswa nokusebenzisa zonke izici ezivamile eziyi-10 (0.8999).

Isiphetho:

Ukusebenza kwe-MemTrax kungasetshenziswa ngempumelelo kumodeli yokuqagela yokuhlukanisa umshini wokufunda uhlelo lokusebenza lokuhlola ukuthola ukukhubazeka kwengqondo kwesigaba sokuqala.

ISINGENISO

Izehlakalo ezibonwayo (nakuba zingaxilongwanga kahle) ezisabalele kanye nokusabalala kanye nokukhula okuhambisanayo kwezokwelapha, kwezenhlalo, kanye nomphakathi zezempilo Izindleko kanye nomthwalo wesifo i-Alzheimer's (AD) kanye nokukhubazeka kwengqondo okuncane (MCI) kuya ngokuya kuba nzima kubo bonke ababambiqhaza [1, 2]. Lesi simo esidabukisayo nesicindezelayo sesenze isicelo esiphuthumayo sokuthi ucwaningo luqinisekise ukutholwa kusenesikhathi ukuhlolwa kwengqondo kanye namathuluzi okuhlola okusetshenziswa okujwayelekile okusebenzayo kuzilungiselelo zomuntu siqu nezomtholampilo ezigulini esezikhulile kuzo zonke izifunda ezihlukene nabantu [3]. Lawa mathuluzi kufanele futhi ahlinzeke ngokuhumusha okungenazihibe kwemiphumela enolwazi kumarekhodi ezempilo e-elekthronikhi. Izinzuzo zizotholakala ngokwazisa iziguli kanye nokusiza odokotela ekuboneni izinguquko ezibalulekile ngaphambili futhi ngaleyo ndlela kunikeze amandla ukuhlukaniswa okusheshayo nokufika ngesikhathi, ukuqaliswa, nokulandelelwa kokwelashwa okufanelekile komuntu ngamunye futhi okungabizi kakhulu kanye nokunakekelwa kwesiguli kulabo abaqala ukuzwa. ukwehla kwengqondo [3, 4].

Ithuluzi lekhompyutha le-MemTrax (https://memtrax.com) ukuhlola kokuqashelwa okulula nokufushane okuqhubekayo okungalawulwa ngokwakho ku-inthanethi ukuze kulinganise ukusebenza kwenkumbulo yesiqephu esinesikhathi esiyinselele lapho umsebenzisi ephendula izithombe eziphindaphindiwe hhayi isethulo sokuqala [5, 6]. Ucwaningo lwakamuva kanye nemithelela esebenzayo ewumphumela iqala ukukhombisa ngokuqhubekayo nangokuhlangene ukusebenza komtholampilo kwe-MemTrax ekuqaleni kokuhlolwa kwe-AD kanye ne-MCI [5-7]. Nokho, ukuqhathanisa okuqondile kokusetshenziswa komtholampilo nokukhona impilo yomqondo ukuhlola kanye nezindinganiso ezivamile kugunyazwe ukwazisa umbono wochwepheshe futhi kuqinise insiza ye-MemTrax ekutholakaleni kusenesikhathi kanye nokwesekwa kokuxilonga. van der Hoek et al. [8] uqhathanise amamethrikhi okusebenza e-MemTrax akhethiwe (isivinini sokusabela namaphesenti alungile) nesimo somqondo njengoba kunqunywa iMontreal Ukuhlola ingqondo (MoCA). Nokho, lolu cwaningo belukhawulelwe ukuhlobanisa lawa mamethrikhi okusebenza nokuchazwa kwesimo somqondo (njengoba kunqunywa i-MoCA) nokuchaza ububanzi obuhlobene namanani anqamule. Ngakho-ke, ukuze sandise lolu phenyo futhi sithuthukise ukusebenza kwezigaba nokusebenza ngempumelelo, umbuzo wethu oyinhloko wocwaningo bekungukuthi:

  • Ingakwazi yini amamethrikhi okusebenza e-MemTrax akhethiwe kanye nezibalo zabantu nempilo efanele Iphrofayili izici kufanele zisetshenziswe ngempumelelo kumodeli yokubikezela ethuthukiswe ngokufunda komshini ukuze kuhlukaniswe impilo yengqondo ngendlela ehlukile (evamile uma iqhathaniswa ne-MCI), njengoba kungaboniswa umphumela womuntu we-MoCA?

Okwesibili kulokhu, besifuna ukwazi:

  • Kubandakanya izici ezifanayo, ingabe imodeli yokufunda yomshini esekelwe ekusebenzeni kwe-MemTrax ingasetshenziswa ngokuphumelelayo esigulini ukuze sibikezele ubukhali (obumnene uma buqhathaniswa nobukhulu) phakathi kwezigaba ezikhethiwe zokukhubazeka kwengqondo njengoba kunganqunywa ukuxilongwa okuzimele komtholampilo?

Ukuvela nokuthuthuka kokusetshenziswa kobuhlakani bokwenziwa nokufunda komshini ekuhloleni/kutholwa sekuvele kubonise izinzuzo ezingokoqobo ezihlukile, ngokumodela okubikezelwayo okuqondisa ngempumelelo odokotela ekuhloleni okuyinselele kwengqondo/ubuchopho kanye nokuphathwa kwesiguli. Ocwaningweni lwethu, sikhethe indlela efanayo ekufanisweni kwezigaba ze-MCI kanye nokubandlululwa kobunzima bokukhubazeka kwengqondo njengoba kuqinisekiswa ukuxilongwa komtholampilo okuvela kumasethi wedatha amathathu amele iziguli ezilaliswa ngokuzithandela ezikhethiwe kanye neziguli zangaphandle ezivela ezibhedlela ezimbili e-China. Sisebenzisa imodeli yokubikezela ukufundwa komshini, sihlonze abafundi abenza kahle kakhulu kusukela ezihlanganisweni ezihlukahlukene zedathasethi/zabafundi futhi saklelisa izici ezizosiqondisa ekuchazeni amamodeli asebenza kakhulu ezempilo.

Imibono yethu ibiwukuthi imodeli eqinisekisiwe esekelwe ku-MemTrax ingasetshenziswa ukuhlukanisa impilo yengqondo ngendlela ehlukile (evamile noma i-MCI) ngokusekelwe kumbandela womkhawulo wamaphuzu we-MoCA, kanye nokuthi imodeli efanayo yokuqagela ye-MemTrax ingasetshenziswa ngempumelelo ekubandlululeni ubukhali ezigabeni ezikhethiwe ze kuxilongwa ngomtholampilo ukungakhuli engqondweni yethu. Ukubonisa imiphumela elindelwe kungaba usizo ekusekeleni ukusebenza kahle kwe-MemTrax njengesikrini sokutholwa kusenesikhathi sokuncipha kwengqondo nokuhlukaniswa kokuphazamiseka kwengqondo. Ukuqhathanisa okuhle nezinga okuhloswe ngalo ukuthi lihambisana nokulula okukhulu nokushesha kokusetshenziswa kungaba nomthelela ekusizeni odokotela basebenzise leli thuluzi elilula, elithembekile, nelifinyelelekayo njengesikrini sokuqala ekutholeni ukushoda kwengqondo kwesiteji kusenesikhathi (okuhlanganisa ne-prodromal). Indlela enjalo kanye nokusebenziseka ngaleyo ndlela kungabangela ukunakekelwa nokungenelela kwesiguli okufika ngesikhathi futhi okuhlukaniswe kangcono. Le mininingwane yokucabanga phambili kanye namamethrikhi namamodeli athuthukisiwe kungase futhi kube usizo ekwehliseni noma ekumiseni ukuqhubeka kokuwohloka komqondo, okuhlanganisa i-AD kanye nokuwohloka komqondo okuhlobene ne-AD (ADRD).

IZIMPAHLA NEZINDLELA

Isifundo sabantu

Phakathi kukaJanuwari 2018 no-Agasti 2019, ucwaningo oluhlukene lwaqedwa ngeziguli ezibuthwe ezibhedlela ezimbili zaseChina. Ukuphathwa kwe-MemTrax [5] kubantu abaneminyaka engu-21 nangaphezulu kanye nokuqoqwa nokuhlaziywa kwaleyo datha kwabuyekezwa futhi kwagunyazwa futhi kwalawulwa ngokuvumelana nezindinganiso zokuziphatha ze Human IKomidi Lokuvikela Isihloko laseStanford University. I-MemTrax kanye nakho konke okunye ukuhlolwa kwalolu cwaningo lulonke kwenziwa ngokwesimemezelo sase-Helsinki sango-1975 futhi kwagunyazwa Ibhodi Lokubuyekeza Isikhungo Sesibhedlela Esihlanganisiwe Sokuqala sase-Kunming Medical University e-Kunming, e-Yunnan, e-China. Umsebenzisi ngamunye unikezwe i imvume enolwazi funda/ukubukeza bese uvuma ngokuzithandela ukubamba iqhaza.

Abahlanganyeli baqashwe echibini leziguli ezingaphandle emtholampilo wezinzwa esibhedlela sase-Yanhua (i-YH sub-dataset) kanye umtholampilo wenkumbulo eSibhedlela Esihlanganisiwe Sokuqala saseKunming Medical Inyuvesi (i-XL sub-dataset) ese-Beijing, e-China. Ababambiqhaza baphinde baqashwa kusukela ku-neurology (i-XL sub-dataset) kanye nemithi yangaphakathi (i-KM sub-dataset) ezigulini ezilaliswayo e-First Affiliated Hospital of Kunming Medical University. Imibandela yokufakwa yayihlanganisa 1) amadoda nabesifazane okungenani abaneminyaka engu-21 ubudala, 2) ikhono lokukhuluma isiShayina (Mandarin), kanye 3) nekhono lokuqonda iziqondiso zomlomo nezibhaliwe. Imibandela yokungafakwa kwaba wukubona kanye nokukhubazeka kwezimoto okuvimbela ababambiqhaza ukuthi baqedele Ukuhlolwa kwe-MemTrax, kanye nokungakwazi ukuqonda imiyalelo ethile yokuhlola.

Inguqulo yesiShayina ye-MemTrax

I-intanethi Inkundla yokuhlola ye-MemTrax yahunyushwa kusiShayina (URL: https://www.memtrax.com.cn) futhi sashintshwa ukuze sisetshenziswe nge-WeChat (Shenzhen Tencent Computer Systems Co. LTD., Shenzhen, Guangdong, China) ukuze uzilawule. Idatha igcinwe kuseva yefu (i-Ali Cloud) etholakala e-China futhi inikezwe ilayisense yakwa-Alibaba (Alibaba Technology Co. Ltd., Hangzhou, Zhejiang, China) yi-SJN Biomed LTD (Kunming, Yunnan, China). Imininingwane eqondile ku-MemTrax kanye nemibandela yokuqinisekisa yokuhlola esetshenziswe lapha ichazwe ngaphambilini [6]. Ukuhlolwa kwanikezwa mahhala ezigulini.

Izinqubo zokutadisha

Ezigulini ezilaliswayo kanye neziguli ezilaliswayo, uhlu lwemibuzo olujwayelekile lwephepha lokuqoqa imininingwane yabantu kanye nolwazi lomuntu siqu njengeminyaka, ubulili, iminyaka yemfundo, umsebenzi, uhlala wedwa noma nomndeni, futhi umlando wezokwelapha waphathwa yilungu lethimba locwaningo. Ngemva kokuphothulwa kohlu lwemibuzo, ukuhlolwa kwe-MoCA [12] kanye ne-MemTrax kwasetshenziswa (i-MoCA kuqala) kungabi ngaphezu kwemizuzu engama-20 phakathi kokuhlolwa. Iphesenti le-MemTrax elilungile (MTx-% C), isikhathi sokuphendula esiqondile (MTx-RT), kanye nosuku nesikhathi sokuhlolwa kwarekhodwa ephepheni yilungu lethimba locwaningo kumhlanganyeli ngamunye ohloliwe. Uhlu lwemibuzo olugcwalisiwe kanye nemiphumela ye-MoCA kwalayishwa kusipredishithi se-Excel ngumcwaningi ophethe izivivinyo futhi waqinisekiswa uzakwabo ngaphambi kokuba amafayela e-Excel agcinwe ukuze ahlaziywe.

Ukuhlolwa kwe-MemTrax

Ukuhlolwa okuku-inthanethi kwe-MemTrax kwakuhlanganisa izithombe ezingu-50 (eziyingqayizivele ezingu-25 nezimpinda ezingu-25; amasethi angu-5 ezithombe ezi-5 zezigcawu ezivamile noma izinto) eziboniswa ngohlelo oluthile olungahleliwe. Umbambi qhaza (ngokuyalelwa) angathinta inkinobho ethi Qala esikrinini ukuze aqale ukuhlola futhi aqale ukubuka uchungechunge lwezithombe aphinde athinte isithombe esisesikrinini ngokushesha okukhulu noma nini lapho kuvela isithombe esiphindaphindiwe. Isithombe ngasinye sivele amasekhondi angu-3 noma kwaze kwathintwa isithombe esikusikrini, okwadala ukwethulwa kwesithombe esilandelayo ngokushesha. Kusetshenziswa iwashi langaphakathi ledivayisi yasendaweni, i-MTx-RT yesithombe ngasinye yanqunywa isikhathi esidlulile kusukela ekubonisweni kwesithombe kuya lapho isikrini sithintwa umhlanganyeli ekuphenduleni ekuboniseni ukubonwa kwesithombe njengesivele sibonisiwe. ngesikhathi sokuhlolwa. I-MTx-RT irekhodwe kuso sonke isithombe, ngama-3 agcwele aqoshiwe angabonisi mpendulo. I-MTx-% C ibalwe ukuze kuboniswe iphesenti lokuphinda nezithombe zokuqala lapho umsebenzisi aphendule khona ngendlela efanele (okuphozithivu kweqiniso + okunegethivu kweqiniso kuhlukaniswe ngo-50). Imininingwane eyengeziwe yokuphathwa nokuqaliswa kwe-MemTrax, ukuncishiswa kwedatha, idatha engavumelekile noma "ayikho impendulo", kanye nokuhlaziywa kwedatha okuyinhloko kuchazwe kwenye indawo [6].

Ukuhlolwa kwe-MemTrax kwachazwa ngokuningiliziwe futhi ukuhlolwa kokuprakthiza (okunezithombe eziyingqayizivele ngaphandle kwalezo ezisetshenziswe ekuhlolweni ukuze kuqoshwe imiphumela) kwanikezwa ababambiqhaza esimweni sasesibhedlela. Abahlanganyeli kudathasethi engaphansi ye-YH ne-KM bathathe isivivinyo se-MemTrax ku-smartphone eyayilayishwe uhlelo lokusebenza ku-WeChat; kuyilapho inani elilinganiselwe leziguli zedathasethi engaphansi ye-XL zisebenzisa i-iPad kanti ezinye zisebenzisa i-smartphone. Bonke ababambiqhaza bathathe isivivinyo se-MemTrax umphenyi wocwaningo ebhekile ngokunganaki.

Ukuhlolwa kwengqondo kweMontreal

Inguqulo yaseBeijing ye-Chinese MoCA (MoCA-BC) [13] yaphathwa futhi yatholwa abacwaningi abaqeqeshiwe ngokwemiyalelo yokuhlola esemthethweni. Ngokufanelekile, i-MoCA-BC ikhonjiswe ukuthi ithembekile ukuhlolwa kwengqondo ukuhlolwa kuwo wonke amazinga emfundo kubantu abadala baseShayina asebekhulile [14]. Ukuhlolwa ngakunye kuthathe cishe imizuzu eyi-10 ukuya kwengama-30 ukulawulwa ngokusekelwe kumakhono okuqonda ombambi qhaza abafanele.

Imodeli yokuhlukaniswa kwe-MoCA

Kube nengqikithi yezinto ezisebenzisekayo ezingama-29, okuhlanganisa ne-MemTrax emibili amamethrikhi okusebenza kokuhlola nezici ezingu-27 ezihlobene nezibalo nempilo ulwazi lomhlanganyeli ngamunye. Isilinganiso sesivivinyo se-MoCA sesiguli ngasinye sisetshenziswe njenge ukuhlolwa kwengqondo "ibhentshimark" yokuqeqesha amamodeli ethu aqagelayo. Ngokufanelekile, ngenxa yokuthi i-MoCA yasetshenziswa ukudala ilebula yekilasi, asikwazanga ukusebenzisa amaphuzu ahlanganisiwe (noma amanye amaphuzu esethi engaphansi ye-MoCA) njengesici esizimele. Senze ukuhlola kokuqala lapho senze khona imodeli (sihlukanisa impilo yengqondo echazwe yi-MoCA) amasethi wedatha angaphansi kwesibhedlela/imitholampilo amathathu ngawodwana bese sihlanganiswa sisebenzisa zonke izici. Nokho, zonke izici ezifanayo zedatha aziqoqwanga kumtholampilo ngamunye kwemine emele amasethi amancane wedatha; ngakho, izici zethu eziningi kudathasethi ehlanganisiwe (uma kusetshenziswa zonke izici) zinezehlakalo eziphezulu zamanani angekho. Sibe sesakha amamodeli anedathasethi ehlanganisiwe sisebenzisa izici ezivamile kuphela ezibangele ukusebenza kwezigaba okuthuthukisiwe. Lokhu cishe kwachazwa inhlanganisela yokuba nezimo eziningi zokusebenza ngokuhlanganisa amasethi edatha angaphansi kweziguli ezintathu futhi azikho izici ezinobuningi obungadingekile bamanani angekho (isici esisodwa kuphela kudathasethi ehlanganisiwe, uhlobo lomsebenzi, esinamanani ashodayo, athinta izehlakalo ezintathu kuphela zesiguli), ngoba izici ezijwayelekile kuphela ezirekhodwe kuzo zontathu izizinda ezifakiwe. Ngokuphawulekayo, besingenawo umbandela othile wokwenqaba isici ngasinye esingazange sifakwe kudathasethi ehlanganisiwe. Kodwa-ke, ekufanisweni kwethu kokuqala kwedathasethi ehlanganisiwe, siqale sasebenzisa zonke izici ezivela kudathasethi engaphansi kwesiguli ngasinye kwezintathu ezihlukene. Lokhu kubangele kabanzi ekusebenzeni kwemodeli okwakungaphansi ngokulinganiswa kunomodeli wokuqala wokuqala kudathasethi ngayinye encane. Ngaphezu kwalokho, nakuba ukusebenza ngezigaba kwamamodeli akhiwe kusetshenziswa zonke izici bekukhuthaza, kubo bonke abafundi nezikimu zokuhlukanisa, ukusebenza kwaba ngcono ngamamodeli amaningi ngokuphindwe kabili uma kusetshenziswa izici ezivamile kuphela. Eqinisweni, phakathi kwalokho okwagcina kube ngabafundi bethu abaphambili, zonke ngaphandle kwemodeli eyodwa zaba ngcono ekuqedeni izici ezingajwayelekile.

Isethi yedatha ehlanganisiwe yokugcina (i-YH, i-XL, ne-KM kuhlangene) ihlanganisa izehlakalo ezingu-259, ngayinye imele umhlanganyeli oyingqayizivele othathe kokubili ukuhlolwa kwe-MemTrax ne-MoCA. Kube nezici ezizimele ezingu-10 ezabiwe: Amamethrikhi okusebenza kwe-MemTrax: MTx-% C kanye ne-MTx-RT; imininingwane yomlando wezibalo zabantu kanye nowezokwelapha: iminyaka, ubulili, iminyaka yokufunda, uhlobo lomsebenzi (ikhola eluhlaza okwesibhakabhaka/ikholomu emhlophe), ukwesekwa komphakathi (ukuthi ohlolayo uhlala yedwa noma nomndeni), kanye nezimpendulo zikayebo/cha zokuthi umsebenzisi ubenayo yini umlando wesifo sikashukela, i-hyperlipidemia, noma ukulimala kobuchopho okubuhlungu. Amamethrikhi amabili engeziwe, i-aggregate score ye-MoCA kanye ne-MoCA aggregate score elungiselwe iminyaka yokufunda [12], asetshenziswe ngokuhlukana ukuze kuthuthukiswe amalebula okuhlukaniswa okuncikile, ngaleyo ndlela kwakha izikimu zokumodela ezimbili ezihlukene ezizosetshenziswa kudathasethi yethu ehlanganisiwe. Enguqulweni ngayinye (elungisiwe nengalungiswanga) yesikolo se-MoCA, idatha iphinde yamodela ngokuhlukene ukuze ihlukaniswe kanambambili kusetshenziswa imibandela emibili ehlukene—owokuqala onconyiwe [12] kanye nelinye inani elisetshenziswa futhi lakhuthazwa abanye [8, 15]. Kwesinye isikimu sokuhlukanisa i-threshold, isiguli sasibhekwa njengengqondo evamile uma sithole u-≥23 ekuhlolweni kwe-MoCA futhi sine-MCI uma isikolo sasingu-22 noma ngaphansi; kanti, kufomethi yokuhlukanisa enconyiwe yokuqala, isiguli kwadingeka sithole amaphuzu angama-26 noma angcono ku-MoCA ukuze sibhalwe ukuthi sinempilo evamile yokucabanga.

Idatha ehlungiwe yokumodeliswa kwezigaba ze-MoCA

Siphinde sahlola ukuhlukaniswa kwe-MoCA sisebenzisa izindlela ezine zokukala izici ezivame ukusetshenziswa: i-Chi-Squared, i-Gain Ratio, i-Information Gain, kanye ne-Symmetrical Uncertainty. Ngombono wesikhashana, sisebenzise abalinganisi kuyo yonke idathasethi ehlanganisiwe sisebenzisa izikimu zethu zokumodela ezine. Bonke abaphathi bavumelene ngezici ezifanayo eziphezulu, okungukuthi, ubudala, inombolo yeminyaka yemfundo, kanye nawo womabili amamethrikhi okusebenza e-MemTrax (MTx-% C, okusho ukuthi MTx-RT). Sibe sesakha kabusha amamodeli sisebenzisa inqubo yokukhetha isici ngasinye ukuze siqeqeshe amamodeli ngezici ezine eziphezulu kuphela (bona Ukukhetha kwesici ngezansi).

Imiphumela yokwehluka kokugcina eyisishiyagalombili yezinhlelo ze-MoCA zokuhlukanisa amaphuzu zethulwe kuThebula 1.

Ithebula 1

Isifinyezo sokuhluka kwesikimu sokumodela esisetshenziselwa ukuhlukaniswa kwe-MoCA (Okuvamile Impilo Yengqondo ngokumelene ne-MCI)

Uhlelo LokumodelaImpilo Ejwayelekile Yokuqonda (Ikilasi Elibi)I-MCI (Ikilasi Elihle)
Ilungisiwe-23 Ayihlungiwe/Ayihlungiwe101 (39.0%)158 (61.0%)
Ilungisiwe-26 Ayihlungiwe/Ayihlungiwe49 (18.9%)210 (81.1%)
Okungalungiswanga-23 Akuhlungiwe/Akuhlungiwe92 (35.5%)167 (64.5%)
Okungalungiswanga-26 Akuhlungiwe/Akuhlungiwe42 (16.2%)217 (83.8%)

Inombolo efanele kanye namaphesenti engqikithi yeziguli ekilasini ngalinye ahlukaniswa ngokulungiswa kwamaphuzu emfundo (Alungisiwe noma Angalungisiwe) kanye nomkhawulo wokuhlukaniswa (23 noma 26), njengoba kusetshenziswe kuzo zombili amasethi ezici (Azihlungiwe futhi Ezihlungiwe).

I-MemTrax-based based evaluation modelling

Kumadathasethi ethu amathathu angaphansi (YH, XL, KM), yiziguli zedathasethi engaphansi ye-XL kuphela ezaxilongwa ngokuzimela ngokukhubazeka kwengqondo (okungukuthi, amaphuzu azo e-MoCA ahlukene awazange asetshenziswe ekusunguleni ukuhlukaniswa kwezigaba ezivamile nezikhubazekile). Ngokukhethekile, iziguli ze-XL zatholakala ukuthi zinazo noma Ukuhlolwa kwesifo i-Alzheimer's (AD) noma ukuwohloka komqondo kwemithambo (VaD). Ngaphakathi kwalezi zigaba zokuxilongwa okuyinhloko, kwaba khona okunye ukuqokwa kwe-MCI. Ukuxilongwa kwe-MCI, ukuwohloka komqondo, i-vascular neurocognitive disorder, kanye ne-neurocognitive disorder ngenxa ye-AD kwakusekelwe ezimisweni zokuxilonga eziqondile nezihlukile ezichazwe ku-Diagnostic and Statistical Manual of Mental Disorders: DSM-5 [16]. Uma kucatshangelwa lokhu kuxilongwa okucolisisiwe, izikimu ezimbili zokumodela zezigaba zasetshenziswa ngokuhlukene kudathasethi engaphansi ye-XL ukuze kuhlukaniswe izinga lobunzima (izinga lokonakala) esigabeni ngasinye sokuxilongwa esiyinhloko. Idatha esetshenziswe kuhlelo ngalunye lwalezi zikimu zokumodela zokuxilonga (AD ne-VaD) yayihlanganisa ulwazi lwezibalo zabantu kanye nomlando wesiguli, kanye nokusebenza kwe-MemTrax (MTx-% C, kusho i-MTx-RT). Ukuxilongwa ngakunye kwakubhalwe ukuthi mnene uma kuqokiwe i-MCI; ngaphandle kwalokho, kwakubhekwa njengento enzima. Siqale sacubungula ukufaka phakathi amaphuzu e-MoCA kumamodeli okuxilongwa (okuncane uma kuqhathaniswa nokuqina); kodwa sinqume ukuthi lokho kuzohlula inhloso yesikimu sethu sokumodela sesibili sokubikezela. Lapha abafundi bazoqeqeshwa kusetshenziswa ezinye izici zesiguli ezitholakala kalula kumhlinzeki kanye namamethrikhi okusebenza ohlolo olulula lwe-MemTrax (esikhundleni se-MoCA) ngokumelene nereferensi “izinga legolide”, ukuxilongwa okuzimele komtholampilo. Kube nezimo ezingama-69 kudathasethi yokuxilongwa kwe-AD kanye nezimo ezingama-76 ze-VaD (Ithebula 2). Kuwo womabili amasethi edatha, bekunezici ezizimele eziyi-12. Ngaphezu kwezici ze-10 ezifakwe kusigaba samaphuzu we-MoCA, umlando wesiguli wawuhlanganisa nolwazi ngomlando we-hypertension kanye nesifo sohlangothi.

Ithebula 2

Isifinyezo sokuhluka kwesikimu sokumodela esisetshenziselwa ukuhlukaniswa kokuqina kokuxilongwa (Kumaphakathi kuqhathaniswa Kunzima)

Uhlelo LokumodelaOkumaphakathi (Ikilasi Elibi)Kunzima (Ikilasi Elihle)
I-MCI-AD ngokumelene ne-AD12 (17.4%)57 (82.6%)
I-MCI-VaD ngokumelene ne-VaD38 (50.0%)38 (50.0%)

Inombolo efanele kanye namaphesenti eziguli eziphelele ekilasini ngalinye zihlukaniswa ngesigaba sokuqala sokuxilongwa (AD noma i-VaD).

Izibalo

Ukuqhathaniswa kwezici zombambiqhaza nezinye izici zezinombolo phakathi kwedathasethi engaphansi yesu ngalinye lokuhlukanisa imodeli (ukubikezela impilo yengqondo ye-MoCA nokuqina kokuxilongwa) kwenziwa kusetshenziswa ulimi lohlelo lwePython (inguqulo 2.7.1) [17]. Umehluko wokusebenza kwemodeli ekuqaleni wanqunywa kusetshenziswa isici esisodwa noma ezimbili (njengoba kufanele) i-ANOVA enesikhathi sokuzethemba esingu-95% kanye nokuhlolwa komehluko obalulekile we-Tukey (HSD) ukuze kuqhathaniswe izindlela zokusebenza. Lokhu kuhlolwa komehluko phakathi kokusebenza kwemodeli kwenziwa kusetshenziswa inhlanganisela yePython ne-R (inguqulo 3.5.1) [18]. Sisebenzise le ndlela (yize, ngokungangabazeki ingaphansi kokulungile) njengendlela yosizo lwe-heuristic kulokhu. ekuqaleni ukuze kuqhathaniswe imodeli yokusebenza yokuqala ekulindeleni ukusetshenziswa komtholampilo okungaba khona. Sibe sesisebenzisa ukuhlolwa kwezinga le-Bayesian esayiniwe sisebenzisa ukusatshalaliswa kwangemuva ukuze sinqume amathuba omehluko wokusebenza kwemodeli [19]. Kulokhu kuhlaziya, sisebenzise isikhawu –0.01, 0.01, okubonisa ukuthi uma amaqembu amabili anomehluko wokusebenza ongaphansi kuka-0.01, ayebhekwa afanayo (ngaphakathi kwesifunda sokulingana okungokoqobo), noma kungenjalo ayehlukile (elilodwa elingcono kune omunye). Ukuze senze ukuqhathanisa kwe-Bayesian kwabahlukanisi bezigaba futhi sibale lawa mathuba, sisebenzise ilabhulali ye-baycomp (inguqulo 1.0.2) yePython 3.6.4.

Ukumodela okuqagelayo

Sakhe amamodeli abikezelayo sisebenzisa isamba esihlukile esiyishumi sezinhlelo zethu zokumodela ukuze sibikezele (ukwehlukanisa) umphumela wokuhlolwa kwe-MoCA yesiguli ngasinye noma ubucayi bokuxilongwa komtholampilo. Bonke abafundi basetshenziswa futhi amamodeli akhiwa kusetshenziswa inkundla yesoftware yomthombo ovulekile we-Weka [20]. Ukuhlaziya kwethu kokuqala, sisebenzise izindlela zokufunda eziyi-10 ezivame ukusetshenziswa: Omakhelwane Abangu-5 Abaseduze, izinguqulo ezimbili zesihlahla sesinqumo se-C4.5, i-Logistic Regression, i-Multilayer Perceptron, i-Naïve Bayes, izinguqulo ezimbili ze-Random Forest, i-Radial Basis Function Network, ne-Support Vector. Umshini. Izibaluli eziyinhloko nokugqama kwalezi zindlela zokuziphatha kuchazwe kwenye indawo [21] (bona Isithasiselo esilandelanayo). Lezi zikhethwe ngenxa yokuthi zimele izinhlobo ezahlukene zabafundi nangenxa yokuthi sibonise impumelelo ngokuzisebenzisa ekuhlaziyeni kwangaphambilini kwedatha efanayo. Izilungiselelo zepharamitha ye-Hyper zikhethwe ocwaningweni lwethu lwangaphambilini ezibonisa ukuthi ziqinile kudatha ehlukahlukene ehlukahlukene [22]. Ngokusekelwe emiphumeleni yokuhlaziya kwethu kokuqala kusetshenziswa idathasethi efanayo ehlanganisiwe enezici ezifanayo eziye zasetshenziswa kamuva ekuhlaziyeni okuphelele, sihlonze abafundi abathathu abanikeze ukusebenza okuqinile okungaguquki kuzo zonke izigaba: Ukwehla Kwezinto Ezisetshenziswayo, i-Naïve Bayes, kanye Nomshini We-Vector Yokusekela.

Ukuqinisekisa okuphambanayo kanye nemethrikhi yokusebenza kwemodeli

Kuwo wonke amamodeli abikezelwayo (okuhlanganisa nokuhlaziya kokuqala), imodeli ngayinye yakhiwa kusetshenziswa ukuqinisekiswa okuphambana okuphindwe izikhathi ezingu-10, futhi ukusebenza kwemodeli kukalwa kusetshenziswa Indawo Engaphansi Kwe-Receiver Operating Characteristic Curve (AUC). Ukuqinisekisa okuphambanayo kwaqala ngokuhlukanisa ngokungahleliwe isethi yedatha yesikimu sokumodela ngasinye kweziyi-10 zibe izingxenye ezilinganayo (ukugoqa), kusetshenziswa eziyisishiyagalolunye zalezi zingxenye ngokulandelana kwazo ukuqeqesha imodeli kanye nengxenye esele yokuhlolwa. Le nqubo iphindwe izikhathi ezingu-10, kusetshenziswa isegimenti ehlukile njengokuhlolwa okusethwe ekuphindaphindweni ngakunye. Imiphumela yabe isihlanganiswa ukuze kubalwe umphumela/ukusebenza kwemodeli yokugcina. Kumfundi ngamunye/inhlanganisela yedathasethi, yonke le nqubo iphindwe izikhathi ezingu-10 idatha ihlukaniswa ngokuhlukile isikhathi ngasinye. Lesi sinyathelo sokugcina sehlise ukuchema, saqinisekisa ukuphindaphindeka, futhi sasiza ekunqumeni ukusebenza kwemodeli kukonke. Sekukonke (kumaphuzu e-MoCA nezikimu zokuhlukaniswa kobunzima bokuxilongwa kuhlangene), amamodeli ayi-10 akhiwe. Lokhu kuhlanganisa amamodeli angahlungiwe angu-6,600 (izikimu zokumodela eziyisi-1,800 ezisetshenziswa kudathasethi×6 abafundi×3 igijima×10 ukugoqa = 10 amamodeli) kanye 1,800 amamodeli ahlungiwe (4,800, izikimu zokumodela ezisetshenziswa kudathasethi×4 yabafundi×3 amasu okukhetha izici×4 ukugijima× 10 ukugoqa = 10 amamodeli).

Ukukhetha kwesici

Kumamodeli ahlungiwe, ukukhethwa kwesici (kusetshenziswa izindlela ezine zokukala izici) kwenziwa phakathi kokuqinisekisa okuphambene. Ekugoqeni ngakunye kwayi-10, njengoba u-10% ohlukile wedathasethi bekuyidatha yokuhlola, izici ezine eziphezulu kuphela ezikhethiwe zesethi ngayinye yedatha yokuqeqeshwa (okungukuthi, amanye amafolda ayisishiyagalolunye, noma ama-90% asele ayo yonke idathasethi) asetshenzisiwe. ukwakha amamodeli. Asikwazanga ukuqinisekisa ukuthi yiziphi izici ezine ezisetshenziswe kumodeli ngayinye, njengoba lolo lwazi alugcinwa noma lwenziwa lutholakale ngaphakathi kwenkundla yokumodela esiyisebenzisile (Weka). Nokho, uma kubhekwa ukungaguquguquki ekukhetheni kwethu kokuqala kwezici eziphezulu lapho amazinga asetshenziswa kuyo yonke idathasethi ehlanganisiwe kanye nokufana okwalandela ekusebenzeni kwamamodeli, lezi zici ezifanayo (iminyaka, iminyaka yemfundo, MTx-% C, kanye nencazelo ye-MTx-RT). ) kungenzeka ukuthi yizona eziphezulu ezine ezisetshenziswa kakhulu ezihambisana nokukhetha kwesici ngaphakathi kwenqubo yokuqinisekisa okuphambene.

IZIPHUMA

Izici zezinombolo zombambi qhaza (okuhlanganisa izikolo ze-MoCA namamethrikhi okusebenza e-MemTrax) yamasethi edatha alandelanayo wesu ngalinye lokuhlukanisa imodeli ukubikezela impilo yengqondo ekhonjiswe yi-MoCA (evamile uma iqhathaniswa ne-MCI) kanye nobucayi bokuxilongwa (okuncane uma kuqhathaniswa nokuqina) kuboniswa kuThebula 3.

Ithebula 3

Izici zombambi qhaza, amaphuzu e-MoCA, nokusebenza kwe-MemTrax kusu ngalinye lokuhlukanisa imodeli

Isu LokuhlukanisaUbudalaImfundoI-MoCA ishintshiweI-MoCA AyikalungiswangaMTx-% CI-MTx-RT
Isigaba se-MoCA61.9 y (13.1)9.6 y (4.6)19.2 (6.5)18.4 (6.7)I-74.8% (15.0)1.4 isekhondi (0.3)
Ukuqina Kokuxilongwa65.6 y (12.1)8.6 y (4.4)16.7 (6.2)15.8 (6.3)I-68.3% (13.8)1.5 isekhondi (0.3)

Amanani abonisiwe (okusho ukuthi, i-SD) ahlukaniswa ngamasu okuhlukanisa okuyimodeli amele idathasethi ehlanganisiwe esetshenziselwa ukubikezela impilo yokuqonda ekhonjiswe yi-MoCA (MCI uma iqhathaniswa nokujwayelekile) kanye nedathasethi engaphansi ye-XL esetshenziselwa ukubikezela ubucayi bokuxilongwa (okuncane uma kuqhathaniswa nokuqina).

Ngenhlanganisela ngayinye yesikolo se-MoCA (esilungisiwe/esingalungiswanga) kanye ne-threshold (26/23), kube nomehluko wezibalo (p = 0.000) ekuqhathanisweni ngakunye ngakubili (impilo yokuqonda evamile iqhathaniswa ne-MCI) yobudala, imfundo, nokusebenza kwe-MemTrax (MTx-% C kanye ne-MTx-RT). Isethi yedatha engaphansi yesiguli ngasinye ekilasini le-MCI elifanele lenhlanganisela ngayinye yayineminyaka engaba ngu-9 kuya kwengu-15 ubudala, yabika cishe iminyaka emihlanu embalwa yemfundo, futhi yayinokusebenza okuncane okuhle kwe-MemTrax kuwo womabili amamethrikhi.

Imiphumela yokubikezela yokusebenza kokumodela yezigaba zamaphuzu e-MoCA kusetshenziswa abafundi abathathu abaphezulu, i-Logistic Regression, i-Naïve Bayes, ne-Support Vector Machine, ikhonjiswe kuThebula lesi-4. Lezi zintathu zikhethwe ngokusekelwe ekusebenzeni kwabafundi okungaguquguquki okuphezulu kakhulu kuwo wonke amamodeli ahlukahlukene. kusetshenziswe kumadathasethi azo zonke izikimu zokumodela. Kudathasethi engahlungiwe nokumodela, inani ngalinye ledatha kuThebula lesi-4 libonisa ukusebenza kwemodeli okusekelwe ku-AUC ngokulandelanayo esuselwe kumamodeli angu-100 (ama-runs angu-10×10 okugoqa) akhelwe inhlanganisela yesikimu somfundi ngamunye/imodeli, ephezulu ngokulandelanayo. umfundi ophumelele okhonjwe ngokugqamile. Nakuba ekubunjweni kwedathasethi ehlungiwe, imiphumela ebikwe kuThebula lesi-4 ibonisa isilinganiso sesilinganiso sokusebenza samamodeli esisuka kumamodeli angu-400 womfundi ngamunye esebenzisa indlela ngayinye yesici (izindlela zokukala izici ezi-4×10 zigijima×10 ukugoqa).

Ithebula 4

Ukusebenza kokuhlukaniswa kwamaphuzu e-Dichotomous MoCA (AUC; 0.0–1.0) kumphumela womfundi ngamunye kwabathathu abenze kahle kuzo zonke izikimu zokumodela

Isethi Yesici EsisetshenzisiweIsikolo se-MoCAI-Cutoff ThresholdUkucindezelwa KwenhlosoNaïve BayesSekela Vector Machine
Okungahlungiwe (izici eziyi-10)Kuguquliwe230.88620.89130.8695
260.89710.92210.9161
Engalungisiwe230.91030.90850.8995
260.88340.91530.8994
Kuhlungiwe (izici ezi-4)Kuguquliwe230.89290.89540.8948
260.91880.92470.9201
Engalungisiwe230.91350.91340.9122
260.91590.92360.9177

Kusetshenziswa ukuhluka kwesethi yesici, isikolo se-MoCA, kanye nomkhawulo wokunqamula amaphuzu we-MoCA, ukusebenza okuphezulu kakhulu kwesikimu ngasinye sokumodela kuboniswa ku- bold (akuhlukile ngokwezibalo kunabo bonke abanye abangekho phakathi bold kumodeli efanele).

Uma kuqhathaniswa abafundi kuzo zonke izinhlanganisela zezinhlobo zamaphuzu e-MoCA kanye nemikhawulo (elungisiwe/engalungiswanga kanye no-23/26, ngokulandelana) kudathasethi engahlungiwe ehlanganisiwe (okungukuthi, kusetshenziswa izici ezivamile eziyi-10), i-Naïve Bayes ngokuvamile ibingumfundi ophumelele kakhulu ngenani eliphelele. ukusebenza kwezigaba kwe-0.9093. Uma kucatshangelwa abafundi abathathu abaphezulu, ukuhlolwa kwezinga elihlobene ne-Bayesian kubonise ukuthi amathuba (Pr) yaseNaïve Bayes ephumelele ukuhlehla kwe-Logistic ibe ngu-99.9%. Ngaphezu kwalokho, phakathi kwe-Naïve Bayes ne-Support Vector Machine, amathuba angama-21.0% okulingana okungokoqobo ekusebenzeni komfundi (ngakho, amathuba angama-79.0% e-Naïve Bayes esebenza kahle kakhulu kunomshini Wokusekela Vector), kuhlanganiswe namathuba angu-0.0% okuthi Umshini Wokusekela Vector osebenza kangcono, ngokulinganiswa. iqinisa inzuzo yokusebenza kwe-Naïve Bayes. Ukuqhathaniswa okuqhubekayo kwenguqulo yamaphuzu e-MoCA kubo bonke abafundi/imibundu kuphakamise inzuzo encane yokusebenza kusetshenziswa amaphuzu we-MoCA angalungisiwe uma kuqhathaniswa nokulungisiwe (0.9027 kuqhathaniswa no-0.8971, ngokulandelana); Pr (akulungisiwe > kulungisiwe) = 0.988). Ngokufanayo, ukuqhathaniswa kwe-cutoff threshold kubo bonke abafundi nezinguqulo zamaphuzu e-MoCA kubonise inzuzo encane yokusebenza ngezigaba kusetshenziswa u-26 njengomkhawulo wokuhlukanisa uma uqhathaniswa no-23 (0.9056 uqhathaniswa no-0.8942, ngokulandelana; Pr (26 > 23) = 0.999). Okokugcina, ukuhlola ukusebenza kwezigaba kwamamodeli kusetshenziswa imiphumela ehlungiwe kuphela (okungukuthi, izici ezine ezisezingeni eliphezulu kuphela), i-Naïve Bayes (0.9143) ngokwezinombolo ibingumfundi owenze kahle kakhulu kuzo zonke izinguqulo/imibundu ye-MoCA. Kodwa-ke, kuzo zonke izindlela zokulinganisa izici ezihlanganisiwe, bonke abafundi abenze kahle benze okufanayo. Ukuhlolwa kwezinga elisayiniwe kwaseBayesia kubonise amathuba angu-100% okulingana okungokoqobo phakathi kwepheya ngalinye labafundi abahlungiwe. Njengedatha engahlungiwe (kusetshenziswa zonke izici ezijwayelekile eziyi-10), kuphinde kwaba nenzuzo yokusebenza kwenguqulo engalungiswanga yesikolo se-MoCA (Pr (akulungisiwe > okulungisiwe) = 1.000), kanye nenzuzo efanayo ehlukile yomkhawulo wokuhlukanisa we-26 (Pr (26 > 23) = 1.000). Ngokuphawulekayo, ukusebenza okumaphakathi komfundi ngamunye kwabathathu abaphezulu kuzo zonke izinguqulo/imikhawulo yamaphuzu e-MoCA kusetshenziswa izici ezine ezisezingeni eliphezulu kuphela kudlule ukusebenza okumaphakathi kwanoma yimuphi umfundi kudatha engahlungiwe. Akumangazi ukuthi ukusebenza ngokwezigaba kwamamodeli ahlungiwe (kusetshenziswa izici ezine ezisezingeni eliphezulu) sekukonke bekuphakeme (0.9119) kunamamodeli angahlungiwe (0.8999), kungakhathaliseki ukuthi amamodeli wendlela yesici aqhathaniswe nalawo mamodeli ahlukene kusetshenziswa zonke eziyi-10 ezivamile. izici. Endleleni ngayinye yokukhetha isici, bekunethuba elingu-100% lenzuzo yokusebenza ngaphezu kwamamodeli angahlungiwe.

Ngeziguli ezicatshangelwa ukuhlukaniswa kokuqina kokuxilongwa kwe-AD, umehluko phakathi kweqembu (MCI-AD vs AD) ngokweminyaka (p = 0.004), imfundo (p = 0.028), isikolo se-MoCA silungisiwe/asilungiswanga (p = 0.000), kanye ne-MTx-% C (p = 0.008) bezibalulekile ngokwezibalo; kanti kuMTx-RT kwakungeyona (p = 0.097). Ngalezo ziguli ezicatshangelwa ukuhlukaniswa kobunzima bokuxilongwa kwe-VaD, umehluko phakathi kweqembu (MCI-VaD vs VaD) wesikolo se-MoCA esilungisiwe/ esingalungiswanga (p = 0.007) kanye ne-MTx-% C (p = 0.026) kanye ne-MTx-RT (p = 0.001) zazibalulekile ngokwezibalo; kanti ngeminyaka (p = 0.511) kanye nemfundo (p = 0.157) bekungekho umehluko obalulekile phakathi kweqembu.

Imiphumela ebikezelayo yokusebenza kokumodela yezigaba zobucayi bokuxilongwa kusetshenziswa abafundi abathathu abakhethwe ngaphambilini, i-Logistic Regression, i-Naïve Bayes, kanye nomshini we-Support Vector, ikhonjisiwe kuThebula lesi-5. Nakuba abafundi abengeziwe abahloliwe babonise ukusebenza okunamandla kancane ngamunye ngamunye ngesigaba esisodwa kwezimbili zokuxilongwa emtholampilo. , abafundi abathathu esibabone njengabathandeka kakhulu ekumodeleni kwethu kwangaphambili banikeze ukusebenza okungaguquki kuzo zombili izikimu zokumodela ezintsha. Uma kuqhathaniswa abafundi kuzo zonke izigaba eziyinhloko zokuxilongwa (AD ne-VaD), awukho umehluko wokusebenza ngokwezigaba ongaguquki phakathi kwabafundi be-MCI-VaD ne-VaD, nakuba Umshini Wokusekela Vector ngokuvamile wawusebenza ngokugqama kakhulu. Ngokufanayo, awukho umehluko obalulekile phakathi kwabafundi be-MCI-AD ngokumelene nesigaba sika-AD, nakuba i-Naïve Bayes (NB) ibe nenzuzo encane yokusebenza kune-Logistic Regression (LR) kanye nobuningi obuncane obuncane kunomshini Wokusekela Vector, okungenzeka kube ngu-61.4% kanye nama-41.7% ngokulandelana. Kuwo womabili amasethi wedatha, kube khona inzuzo yokusebenza iyonke Yomshini Wokusekela Vector (SVM), nge Pr (SVM > LR) = 0.819 kanye Pr (SVM > NB) = 0.934. Ukusebenza kwethu kwezigaba kuwonke kubo bonke abafundi ekubikezeleni ubucayi bokuxilongwa kudathasethi encane ye-XL bekungcono esigabeni sokuxilongwa kwe-VaD uma kuqhathaniswa no-AD (Pr (VAD > AD) = 0.998).

Ithebula 5

Ukusebenza kwesigaba sokuqina komtholampilo kwe-Dichotomous (AUC; 0.0–1.0) imiphumela yomfundi ngamunye kwabathathu abenze kahle kuzo zombili izikimu zokumodela ezihlukene

Uhlelo LokumodelaUkucindezelwa KwenhlosoNaïve BayesSekela Vector Machine
I-MCI-AD ngokumelene ne-AD0.74650.78100.7443
I-MCI-VaD ngokumelene ne-VaD0.80330.80440.8338

Ukusebenza okuphezulu kakhulu kohlelo ngalunye lokumodela kuboniswa ku bold (akufani ngokwezibalo kunezinye ezingangeni bold).

UKUKHULUMA

Ukutholwa kusenesikhathi kwezinguquko empilweni yengqondo kubalulekile usizo olungokoqobo ekuphathweni kwezempilo yomuntu kanye nempilo yomphakathi ngokufanayo. Ngempela, futhi kubaluleke kakhulu ezindaweni zomtholampilo ezigulini emhlabeni wonke. Umgomo okwabelwana ngawo uwukuxwayisa iziguli, abanakekeli, nabahlinzeki kanye nokusheshisa ukwelashwa okufanele futhi okungabizi kakhulu kanye nokunakekelwa kwesikhathi eside kulabo abaqala ukuzwa ukuncipha kwengqondo. Sihlanganisa ama-subsets edatha ethu ezibhedlela/imitholampilo, sihlonze abafundi abathathu abakhethwa kakhulu (ngokugqama okukodwa -Naïve Bayes) ukuze sakhe amamodeli abikezelayo sisebenzisa. Amamethrikhi okusebenza e-MemTrax angahlukanisa ngokuthembekile isimo sezempilo yengqondo dichotomously (impilo yomqondo evamile noma i-MCI) njengoba kuzoboniswa amaphuzu ahlanganisiwe we-MoCA. Ngokuphawulekayo, ukusebenza ngezigaba sekukonke kubo bonke abafundi abathathu kwaba ngcono lapho amamodeli ethu esebenzisa kuphela izici ezine ezisezingeni eliphezulu ezihlanganisa ngokuyinhloko lawa mamethrikhi okusebenza e-MemTrax. Ngaphezu kwalokho, sembule amandla aqinisekisiwe okusebenzisa abafundi abafanayo kanye namamethrikhi okusebenza e-MemTrax esikimini sokuhlukanisa ngezigaba zokuxilonga ukuze kuhlukaniswe ubukhali bezigaba ezimbili zokuxilongwa kokuwohloka komqondo: i-AD ne-VaD.

Ukuhlola inkumbulo iwumgogodla wokutholwa kusenesikhathi kwe-AD [23, 24]. Ngakho-ke, kunethuba ukuthi i-MemTrax ibe yi-inthanethi eyamukelekayo, ehehayo, futhi kulula ukuyisebenzisa. ukuhlolwa kokuhlolwa kwenkumbulo ye-episodic emphakathini jikelele [6]. Ukunemba kokuqashelwa kanye nezikhathi zokuphendula ezivela kulo msebenzi wokusebenza okuqhubekayo kuveza ikakhulukazi ekuhlonzeni ukuwohloka kwangaphambi kwesikhathi nokushintshayo kanye nokushoda okuwumphumela wezinqubo ze-neuroplastic ezihlobene nokufunda, inkumbulo, nokuqonda. Okusho ukuthi, amamodeli alapha asekelwe kakhulu kumamethrikhi okusebenza e-MemTrax azwela futhi maningi amathuba okuthi aveze kalula futhi ngezindleko eziphansi aveze ukushoda kwe-biological neuropathologic phakathi nesigaba sezinguquko se-asymptomatic ngaphambi kokulahlekelwa okukhulu kokusebenza [25]. Ashford et al. ihlolisise amaphethini nokuziphatha kokunemba kwenkumbulo yokuqashelwa kanye nesikhathi sokuphendula kubasebenzisi abaku-inthanethi ababambe iqhaza bodwa nge-MemTrax [6]. Ukuhlonipha ukuthi lokhu kusatshalaliswa kubalulekile ekwenziweni imodeli efanele nasekuthuthukiseni izinhlelo zokusebenza zokunakekelwa kwesiguli ezisebenzayo nezisebenzayo, ukucacisa amaphrofayili asebenzayo emtholampilo kanye nesikhathi sokuphendula kubalulekile ekusunguleni inkomba eyisisekelo ebalulekile yensiza yomtholampilo nocwaningo. Inani elingokoqobo le-MemTrax ekuhlolweni kwe-AD kokukhubazeka kwengqondo kwesigaba sangaphambi kwesikhathi kanye nokwesekwa kokuxilonga okuhlukile lidinga ukube selihlolisiswa kabanzi esimeni somtholampilo lapho kungacatshangelwa khona izimo ezihambisana nakho kanye nekhono lokuqonda, izinzwa, kanye namandla okunyakazisa athinta ukusebenza kokuhlolwa. Futhi ukwazisa umbono wobungcweti nokukhuthaza ukusetshenziswa okusebenzayo komtholampilo, kubalulekile okokuqala ukubonisa ukuqhathanisa nokuhlolwa kokuhlolwa kwezempilo kwengqondo okumisiwe, nakuba lokhu kwakamuva kungase kuvinjwe indlela yokuhlola enzima, izithiyo zolimi nolimi, namathonya amasiko [26] . Mayelana nalokhu, ukuqhathanisa okuhle kwe-MemTrax ekusebenzeni komtholampilo ne-MoCA okuvame ukubizwa ngokuthi izinga lemboni kubalulekile, ikakhulukazi uma kulinganiswa ukusebenziseka kalula okukhulu nokwamukela isiguli i-MemTrax.

Ukuhlola kwangaphambilini okuqhathanisa i-MemTrax ne-MoCA kugqamisa ubufakazi obunengqondo kanye nobufakazi bokuqala obugunyaza uphenyo lwethu lokumodela [8]. Nokho, lokhu kuqhathanisa kwangaphambilini kumane kuhlobanise amamethrikhi amabili abalulekile okusebenza kwe-MemTrax esiwahlole ngesimo somqondo njengoba kunqunywa i-MoCA futhi kwachazwa ububanzi obuhlukahlukene namanani anqamule. Sijule ukuhlolwa kokusebenza komtholampilo kwe-MemTrax ngokuhlola indlela yokubikezela esekelwe ekufaniseni enganikeza ukucatshangelwa komuntu ngamunye kwamanye amapharamitha aqondene nesiguli okungenzeka afanele. Ngokuphambene nabanye, asizange sithole inzuzo ekusebenzeni kwemodeli kusetshenziswa ukulungiswa kwemfundo (ukulungiswa) kumphumela we-MoCA noma ekushintshanisweni komkhawulo wamaphuzu we-MoCA wokucatshangelwa kwengqondo kusukela kokunconywe ekuqaleni kokungu-26 kuya ku-23 [12, 15]. Eqinisweni, inzuzo yokusebenza ngezigaba ithandwa kusetshenziswa isikolo se-MoCA esingalungisiwe kanye nomkhawulo ophezulu.

Amaphuzu abalulekile ekusebenzeni komtholampilo

Ukufunda ngomshini kuvame ukusetshenziswa kangcono kakhulu futhi kusebenza kahle ekumodeleni okubikezelwayo lapho idatha ibanzi futhi ine-dimensional, okungukuthi, lapho kunokubhekwa okuningi kanye nohlu olubanzi oluhambisanayo lwezibaluli zenani eliphezulu (elinikelayo). Nokho, ngale datha yamanje, amamodeli ahlungiwe anezici ezine kuphela ezikhethiwe asebenze kangcono kunalawo asebenzisa zonke izici eziyi-10 ezivamile. Lokhu kuphakamisa ukuthi idathasethi yethu yesibhedlela esihlanganisiwe ibingenazo izici ezifanele ngokomtholampilo (inani eliphezulu) ukuze zihlukanise kahle iziguli ngale ndlela. Noma kunjalo, isici sokugcizelela izinga lokusebenza kwamamethrikhi okhiye wokusebenza kwe-MemTrax—MTx-% C kanye ne-MTx-RT—kusekela ngokuqinile ukwakha amamodeli okuhlola ukushoda kwengqondo okuhambisana nalokhu kuhlolwa okulula, okulula ukukusebenzisa, okunezindleko eziphansi, futhi okuveza ngokufanelekile mayelana nalokhu kuhlolwa. ukusebenza kwenkumbulo, okungenani njengamanje njengesikrini sokuqala sohlelo olunambambili lwesimo sezempilo yengqondo. Njengoba kunikezwe ubunzima obuhlala bukhuphuka kubahlinzeki nezinhlelo zokunakekelwa kwezempilo, izinqubo zokuhlolwa kwesiguli kanye nezicelo zomtholampilo kufanele zithuthukiswe ngokufanelekile kugcizelelwe ekuqoqeni, ekulandeleni, nasekufaniseni lezo zici zesiguli kanye namamethrikhi okuhlola awusizo kakhulu, anenzuzo, futhi afakazelwe asebenza ngempumelelo ekuxilongweni. kanye nokwesekwa kokuphathwa kwesiguli.

Njengoba amamethrikhi amabili ayinhloko e-MemTrax eyinhloko ekuhlukaniseni kwe-MCI, umfundi wethu owenze kahle kakhulu (i-Naïve Bayes) ube nokusebenza kokubikezela okuphezulu kakhulu kumamodeli amaningi (i-AUC engaphezu kuka-0.90) enesilinganiso seqiniso-esiqondile kuya kwamanga esisondela noma esidlula ngandlela-thile ku-4. : 1. Uhlelo lokusebenza lomtholampilo oluhumushayo olusebenzisa lo mfundi ngaleyo ndlela lungathwebula (luhlukanise kahle) iningi lalabo abanokushoda kwengqondo, kuyilapho kunciphisa izindleko ezihambisana nokuhlukanisa ngephutha umuntu onempilo evamile yokuqonda njengonokushoda kwengqondo (okungelona iqiniso) noma ukuntula lokho kuhlelwa kulabo abanokushoda kwengqondo (okungalungile okungamanga). Noma yisiphi salezi zimo zokungahlukaniswa kahle singabeka umthwalo ongafanele ongokwengqondo nongokwenhlalo esigulini nakubanakekeli.

Nakuba ekuhlaziyeni kokuqala nokugcwele sisebenzise bonke abafundi abayishumi ohlelweni ngalunye lokumodela, sigxilise imiphumela yethu kubahlukanisi abathathu ababonisa ukusebenza okuqinile okungaguquki. Lokhu futhi bekuwukugqamisa, ngokusekelwe kule datha, abafundi abebelindeleke ukuthi basebenze ngokwethembela ezingeni eliphezulu ekusetshenzisweni komtholampilo okungokoqobo ekunqumeni ukuhlukaniswa kwesimo somqondo. Ngaphezu kwalokho, ngenxa yokuthi lolu cwaningo beluhloselwe uphenyo oluyisethulo ekusetshenzisweni komshini wokufunda ekuhlolweni kwengqondo kanye nalezi zinselele zomtholampilo ezifika ngesikhathi, senze isinqumo sokugcina amasu okufunda alula futhi enziwa okuvamile, ngokulungiswa kwepharamitha okuncane. Siyabonga ukuthi le ndlela ingahle ikhawulele amandla okuqagela okucaciswe kakhudlwana kwesiguli. Ngokunjalo, nakuba ukuqeqesha amamodeli kusetshenziswa izici eziphezulu kuphela (indlela ehlungiwe) kusazisa ngokuqhubekayo mayelana nale datha (eqondile ekushiyekeni kwedatha eqoqwe futhi igqamisa inani lokuthuthukisa isikhathi somtholampilo esiyigugu nezinsiza), siyaqaphela ukuthi ngaphambi kwesikhathi ukunciphisa ububanzi bamamodeli, ngakho-ke, zonke (kanye nezinye izici) kufanele zicatshangelwe ngocwaningo lwangomuso kuze kube yilapho sinephrofayela ecacile yezici ezibalulekile ezingasebenza kumphakathi obanzi. Ngakho-ke, siphinde siqaphele ngokugcwele ukuthi idatha ebandakanya wonke umuntu futhi emele kabanzi kanye nokuthuthukiswa kwalezi zinhlobo nezinye zingadingeka ngaphambi kokuzihlanganisa esicelweni somtholampilo esiphumelelayo, ikakhulukazi ukwamukela izifo ezithinta ukusebenza kwengqondo ezingadinga ukucatshangelwa ekuhloleni okwengeziwe komtholampilo.

Ukusebenziseka kwe-MemTrax kwabuye kwahlelwa ngokumodela kokuqina kwesifo okusekelwe ekuxilongweni okuhlukene komtholampilo. Ukusebenza kwezigaba okungcono kakhulu ekubikezeleni ubukhali be-VaD (uma kuqhathaniswa ne-AD) bekungenjalo Kuyamangaza uma kunikezwe izici zephrofayili yesiguli kumamodeli aqondene nempilo yemithambo yegazi kanye nobungozi bokushaywa unhlangothi, okungukuthi, umfutho wegazi ophakeme, i-hyperlipidemia, isifo sikashukela, kanye (yebo) umlando wokushaywa unhlangothi. Nakuba bekuyoba okufiseleka kakhulu futhi kufaneleka ukuba kwenziwe ukuhlolwa okufanayo komtholampilo ezigulini ezihambisanayo ezinempilo evamile yokucabanga ukuqeqesha abafundi ngale datha ebandakanya wonke umuntu. Lokhu kufaneleka ngokukhethekile, njengoba i-MemTrax ihloselwe ukusetshenziswa ngokuyinhloko ukuze kutholwe kusenesikhathi ukushoda kwengqondo nokulandelela okulandelayo koshintsho ngalunye. Kuyathandeka futhi ukuthi ukusatshalaliswa kwedatha okufiseleka kakhulu kudathasethi ye-VaD kube negalelo ngokwengxenye ekusebenzeni kwamamodeli okungcono kakhulu uma kuqhathaniswa. I-dataset ye-VaD yayinokulinganisela phakathi kwalezi zigaba ezimbili, kanti isethi yedatha ye-AD eneziguli ze-MCI ezimbalwa kakhulu yayingekho. Ikakhulukazi kumadathasethi amancane, ngisho nezimo ezimbalwa ezengeziwe zingenza umehluko olinganisekayo. Yomibili le mibono iyizimpikiswano ezinengqondo ezisekela umehluko ekusebenzeni kwemodeli yokuqina kwesifo. Kodwa-ke, ukuchasisela ukusebenza okuthuthukisiwe kuzici zezinombolo zesethi yedatha noma izici ezingokwemvelo eziqondene nokwethulwa komtholampilo okucatshangelwayo ngaphambi kwesikhathi. Noma kunjalo, le noveli ibonise ukusetshenziswa kwemodeli yokubikezela ye-MemTrax endimeni yokusekelwa kokuxilonga emtholampilo inikeza umbono obalulekile futhi iqinisekisa ukuphishekela ukuhlolwa okwengeziwe neziguli kulo lonke ukuqhubeka kwe-MCI.

Ukuqaliswa nokusetshenziswa okubonisiwe kwe-MemTrax kanye nalawa mamodeli e-China, lapho ulimi namasiko kuhluke kakhulu kwezinye izindawo ezisetshenziswayo ezisunguliwe (isb., i-France, i-Netherlands, ne-United States) [7, 8, 27], kugcizelela futhi amandla ukuze kwamukelwe umhlaba wonke kanye nenani lomtholampilo leplathifomu esekelwe ku-MemTrax. Lesi isibonelo esingaboniswa ekuzameni ukuvumelanisa idatha kanye nokuthuthukisa izinkambiso zamazwe ngamazwe ezisebenzayo kanye nezinsiza zokumodela zokuhlolwa kwengqondo ezilinganiswe futhi zivumelane kalula nezisetshenziswa emhlabeni wonke.

Izinyathelo ezilandelayo ekwehleni komqondo wokumodela nokusebenzisa

Ukungasebenzi kahle kwengqondo ku-AD kwenzeka ngokuqhubekayo, hhayi ngezigaba noma izinyathelo eziqondile [28, 29]. Kodwa-ke, kulesi sigaba sokuqala, umgomo wethu bekuwukuqala ukusungula ikhono lethu lokwakha imodeli ehlanganisa i-MemTrax engahlukanisa ngokuyisisekelo "okuvamile" kokuthi "okungavamile". Idatha ye-empirical eningi ebandakanyayo (isb., ukuthwebula isithombe sobuchopho, izici zofuzo, ama-biomarker, i-comorbidities, nezimpawu zokusebenza zenkimbinkimbi. imisebenzi edinga ingqondo control) [30] kuzo zonke izifunda zomhlaba ezihlukene, imiphakathi, namaqembu eminyaka yobudala ukuze aqeqeshe futhi athuthuke eyinkimbinkimbi (okuhlanganisa nenhlanganisela efanele enesisindo esifanelekile) amamodeli wokufunda wemishini azosekela izinga elikhulu lokuhlukaniswa okuthuthukisiwe, okungukuthi, amandla okuhlukanisa amaqembu eziguli ezinezigaba ezihlukene. I-MCI ibe amasethi angaphansi amancane nacaca kakhudlwana eceleni kokwehla komqondo okuqhubekayo. Ngaphezu kwalokho, ukuxilonga okuhambisanayo komtholampilo kubantu ngabanye kuzo zonke izifunda zeziguli ezihlukene kubalulekile ukuze qeqesha ngempumelelo lawa mamodeli ahlanganisayo futhi aqine ngokubikezelwa. Lokhu kuzokwenza kube lula ukuphathwa kwamacala ane-stratified okuqondile kulabo abanezizinda ezifanayo, amathonya, kanye namaphrofayili engqondo achazwe kancane futhi ngaleyo ndlela kuthuthukise ukwesekwa kwezinqumo zomtholampilo nokunakekelwa kwesiguli.

Iningi locwaningo lomtholampilo olufanele kuze kube manje lukhulume neziguli okungenani ezinokuwohloka komqondo okuncane; futhi, ngokusebenza, kaningi ukungenelela kwesiguli kuzanywa kuphela ezigabeni ezithuthukile. Kodwa-ke, ngenxa yokuthi ukwehla kwengqondo kuqala ngaphambi kokuba kuhlangatshezwane nemibandela yomtholampilo yokuwohloka komqondo, isikrini sokuqala esisetshenziswe ngempumelelo esisekelwe ku-MemTrax singakhuthaza imfundo efanele yabantu ngabanye ngesifo nokuqhubeka kwaso futhi sisheshe sixazulule ukungenelela okufika ngesikhathi. Ngakho-ke, ukutholwa kusenesikhathi kungasekela ukubandakanyeka okufanele kusukela ekuzivivinyeni, ekudleni, ekusekelweni ngokomzwelo, nasekuthuthukisweni komphakathi kuya ekungeneleleni kwemithi futhi kuqinise izinguquko eziphathelene nesiguli ekuziphatheni nasekuboneni ukuthi ngokukodwa noma ngokuhlangene kunganciphisa noma kumise ukuqhubeka kokuwohloka komqondo [31, 32] . Ngaphezu kwalokho, ngempumelelo ukuhlolwa kusenesikhathi, abantu ngabanye kanye nemindeni yabo bangase batshelwe ukuthi bacabangele izivivinyo zemitholampilo noma bathole ukwelulekwa kanye nokunye ukusekelwa kwezinsizakalo zezenhlalakahle ukuze basize ukucacisa okulindelekile kanye nezinhloso kanye nokuphatha imisebenzi yansuku zonke. Ukuqinisekisa okwengeziwe kanye nokusetshenziswa okusebenzayo okusabalele ngalezi zindlela kungaba wusizo ekunciphiseni noma ekumiseni ukuqhubeka kwe-MCI, AD, ne-ADRD kubantu abaningi.

Ngempela, ukuphela okuphansi kwebanga leminyaka yesiguli ocwaningweni lwethu akumeli inani labantu lokukhathazeka ngokwendabuko nge-AD. Noma kunjalo, isilinganiso seminyaka yeqembu ngalinye esisetshenziswa ezinhlelweni zokumodela ngezigaba ezisekelwe kumphumela we-MoCA/umkhawulo kanye nobunzima bokuxilongwa (Ithebula 3) igcizelela ukuthi iningi elicacile (ngaphezu kwama-80%) lineminyaka okungenani engu-50. Ngakho-ke lokhu kusatshalaliswa kufanele kakhulu ukwenziwa okuvamile, kusekela ukusetshenziswa kwalawa mamodeli emphakathini okubonisa lawo avame ukuthinteka ukuqala kwasekuqaleni kanye nesifo se-neurocognitive esikhulayo ngenxa ye-AD ne-VaD. Futhi, ubufakazi bakamuva kanye nombono kugcizelela lezo zici ezibonwayo (isb., umfutho wegazi ophakeme, ukukhuluphala, isifo sikashukela, nokubhema) ezingase zibe nomthelela ekukhuphukeni kusenesikhathi. Izibalo zengozi yemithambo yabantu abadala kanye ne-midlife kanye nokulimala okucashile kobuchopho bemithambo okwenzeka ngobuqili kube nemiphumela esobala ngisho nasebancane. abantu abadala [33-35]. Ngakho-ke, ithuba lokuqala lokuhlola elingcono kakhulu lokuthola kusenesikhathi isigaba sokushoda kwengqondo kanye nokuqala amasu aphumelelayo okuvimbela kanye nokungenelela ekubhekaneni ngempumelelo nokuwohloka komqondo izovela ekuhloleni izici ezinomthelela kanye nezinkomba ezandulelayo kuwo wonke umkhakha weminyaka, okuhlanganisa ukukhula kwasebuntwaneni okungenzeka ngisho nobungane (kuphawula ukuhlobana kwezinto zofuzo ezifana ne-apolipoprotein E kusukela ekukhulelweni kokuqala).

Empeleni, ukuxilonga okuvumelekile komtholampilo kanye nezinqubo ezibizayo zokuthwebula izithombe ezithuthukisiwe, iphrofayili yofuzo, kanye nezimpawu zokulinganisa ezithembisayo azitholakali kalula noma zingenzeka kubahlinzeki abaningi. Ngakho-ke, ezimweni eziningi, ukuhlukaniswa kwesimo sempilo yengqondo jikelele kungase kudingeke kuthathwe kumamodeli kusetshenziswa amanye amamethrikhi alula anikezwa isiguli (isb., ukuzibika ngokwakho. izinkinga zenkumbulo, imithi yamanje, kanye nemikhawulo yemisebenzi evamile) nezici ezivamile zabantu [7]. Amarejista afana neNyuvesi yaseCalifornia Brain Health Registry (https://www.brainhealthregistry.org/) [27] nezinye ezinobubanzi obukhulu bemvelo bezimpawu zokuzibika, izinyathelo zekhwalithi (isb, ukulala kanye nokuqonda kwansuku zonke), imithi, isimo sezempilo, nomlando, kanye izibalo zabantu ezinemininingwane eminingi zizoba wusizo ekuthuthukiseni nasekuqinisekiseni ukusetshenziswa okungokoqobo kwalawa mamodeli angakudala kakhulu emtholampilo. Ngaphezu kwalokho, ukuhlola okufana ne-MemTrax, ebonise usizo ekuhloleni ukusebenza kwenkumbulo, kungase kunikeze isilinganiso esingcono kakhulu se-AD pathology kunezimpawu zebhayoloji. Njengoba kunikezwe ukuthi isici esiyinhloko se-AD pathology ukuphazamiseka kwe-neuroplasticity kanye nokulahlekelwa okuyinkimbinkimbi kakhulu kwama-synapses, okubonakala njenge-episodic. ukungasebenzi kahle kwenkumbulo, isilinganiso esihlola inkumbulo ye-episodic empeleni kungenzeka hlinzeka ngesilinganiso esingcono somthwalo we-AD pathological kunezimpawu zebhayoloji esigulini esiphilayo [36].

Ngawo wonke amamodeli aqagelayo—noma ahambisana nedatha eyinkimbinkimbi nebandakanyayo evela kubuchwepheshe besimanje kanye nemininingwane yomtholampilo ecolisisiwe kuzo zonke izizinda eziningi noma leyo ekhawulelwe olwazini oluyisisekelo nolwazi olutholakala kalula lwamaphrofayili esiguli akhona—inzuzo eyaziwayo yobuhlakani bokwenziwa. nokufunda komshini ukuthi amamodeli angumphumela angahlanganisa futhi “afunde” ngokungenisa kudatha entsha efanele kanye nombono ohlinzekwa ngokusetshenziswa kohlelo lokusebenza okuqhubekayo. Ukulandela ukudluliswa kobuchwepheshe obusebenzayo, njengoba amamodeli lapha (futhi azothuthukiswa) esetshenziswa futhi athuthukiswe ngamacala amaningi kanye nedatha efanelekile (kuhlanganise neziguli ezinezinkinga ezingase zivele ngokuncipha kwengqondo okulandelayo), ukusebenza kokubikezela kanye nokuhlukaniswa kwezempilo kwengqondo kuzoba namandla kakhulu, okuholela esizeni esisebenza ngempumelelo sokusekela izinqumo zomtholampilo. Lokhu kuguquguquka kuzobonakala ngokugcwele nangokungokoqobo ngokushumeka i-MemTrax esikweni (okuhloswe ngayo amakhono atholakalayo) izinkundla abahlinzeki bezempilo abangazisebenzisa ngesikhathi sangempela emtholampilo.

Okubalulekile ekuqinisekiseni nasekusebenziseni imodeli ye-MemTrax ukuze uthole ukwesekwa kokuxilonga nokunakekelwa kwesiguli yidatha yelongitudinal efunwa kakhulu. Ngokubheka nokurekhoda izinguquko ezihambisanayo (uma zikhona) esimweni somtholampilo kulo lonke uhla olwanele lokujwayelekile ngokusebenzisa i-MCI yesigaba sangaphambi kwesikhathi, amamodeli okuhlolwa okufanele okuqhubekayo nokuhlukaniswa angaqeqeshwa futhi alungiswe njengoba iziguli zikhula futhi zilashwa. Okusho ukuthi, ukusetshenziswa okuphindaphindiwe kungasiza ngokulandelelwa kwesikhathi eside kwezinguquko ezinengqondo ezithambile, ukusebenza kahle kokungenelela, nokugcina ukunakekelwa okunolwazi. Le ndlela ihambisana kakhulu nokusebenza komtholampilo kanye nokuphathwa kwesiguli kanye namacala.

Ukulinganiselwa

Siyayithokozela inselele nokubaluleka kokuqoqa idatha yomtholampilo ehlanzekile emtholampilo/isibhedlela esilawulwayo. Noma kunjalo, bekuyoqinisa ukumodela kwethu ukube amasethi wedatha wethu afaka iziguli eziningi ezinezici ezifanayo. Ngaphezu kwalokho, ngokuqondene nendlela yethu yokuhlonza izifo, bekuyoba okufiseleka kakhulu futhi kufaneleka ukuba kwenziwe ukuhlolwa okufanayo komtholampilo ezigulini ezihambisanayo ezinempilo evamile yokuqonda ukuze kuqeqeshwe abafundi. Futhi njengoba kugcizelelwa ukusebenza kwezigaba okuphezulu kusetshenziswa idathasethi ehlungiwe (kuphela izici ezine ezisezingeni eliphezulu), okuvamile kanye izinyathelo zempilo yomqondo/izinkomba cishe ngabe zithuthukile ukusebenza kokumodela ngenani elikhulu lezici ezivamile kuzo zonke iziguli.

Abanye ababambiqhaza kungenzeka ukuthi ngesikhathi esifanayo babenezinye izifo ezingase zibangele ukushiyeka okudlulayo noma okungapheli kwengqondo. Ngaphandle kwedathasethi engaphansi ye-XL lapho iziguli zazihlukaniswa ngokuxilonga njenge-AD noma i-VaD, idatha ye-comorbidity ayizange iqoqwe/ibikwe echibini lesiguli le-YH, futhi ukugula okubikwe kakhulu kokubi kudathasethi ye-KM kwakuyisifo sikashukela. Kuyaphikiswa, nokho, ukuthi ukufaka iziguli ezinhlelweni zethu zokumodela ezinezinkinga ezingase zibangele noma zandise izinga lokuntula ingqondo kanye nokusebenza okuphansi okungaba umphumela kwe-MemTrax kuzomele kakhulu inani leziguli ezihlosiwe emhlabeni wangempela kulokhu kuhlolwa kokuqonda kwangaphambi kwesikhathi. nendlela yokumodela. Ukuqhubekela phambili, ukuxilongwa okunembile kwe-comorbidities okungase kuthinte ukusebenza kwengqondo kunenzuzo enkulu ekuthuthukiseni amamodeli kanye nezinhlelo zokusebenza zokunakekelwa kwesiguli.

Okokugcina, iziguli zedathasethi engaphansi ye-YH ne-KM zasebenzisa i-smartphone ukuze zihlole i-MemTrax, kuyilapho inani elilinganiselwe leziguli zedathasethi engaphansi ye-XL zisebenzisa i-iPad kanti ezinye zisebenzisa i-smartphone. Lokhu kungenzeka kwethule umehluko omncane ohlobene nedivayisi ekusebenzeni kwe-MemTrax kokumodela kwesigaba se-MoCA. Kodwa-ke, umehluko (uma ukhona) ku-MTx-RT, isibonelo, phakathi kwamadivayisi ngeke ube yinto engekho, ikakhulukazi lapho umhlanganyeli ngamunye enikezwa ukuhlolwa "kokuzilolonga" ngaphambi nje kokusebenza kokuhlola okurekhodiwe. Noma kunjalo, ukusetshenziswa kwalawa madivayisi amabili aphathwa ngesandla kungase kuphazamise ukuqhathaniswa okuqondile kanye/noma ukuhlanganiswa neminye imiphumela ye-MemTrax lapho abasebenzisi baphendule khona ukuphinda izithombe ngokuthinta ibha yesikhala kukhibhodi yekhompyutha.

Amaphuzu angukhiye ku-MemTrax predictive modeling utility

  • • Amamodeli ethu aqagelayo asebenza kahle kakhulu ahlanganisa amamethrikhi okusebenza e-MemTrax akhethiwe angahlukanisa ngokuthembekile isimo sempilo yomqondo (impilo evamile yokuqonda kwengqondo noma i-MCI) njengoba kungaboniswa ukuhlola okwaziwa kabanzi nge-MoCA.
  • • Le miphumela isekela ukuhlanganiswa kwamamethrikhi okusebenza e-MemTrax akhethiwe abe uhlelo lokusebenza lokuhlola imodeli eqagelayo yesigaba sokuqala ukukhubazeka kwengqondo.
  • • Ukumodela kwethu ngokwezigaba kuphinde kwaveza amandla okusebenzisa ukusebenza kwe-MemTrax ezinhlelweni zokuhlukanisa ubunzima bokuxilongwa kokuwohloka komqondo.

Lokhu okutholakele kumanoveli kusungula ubufakazi obuqinisekile obusekela ukusetshenziswa komshini wokufunda ekwakhiweni kwamamodeli okuhlukanisa asekelwe ku-MemTrax aqinile ukuze kusekelwe ukuxilonga ekuphathweni kwecala lomtholampilo okuphumelelayo nokunakekelwa kwesiguli kwabantu abanokukhubazeka kwengqondo.

Ukuvuma

Siyawubona umsebenzi ka-J. Wesson Ashford, Curtis B. Ashford, kanye nozakwabo ngokusungula nokuqinisekisa umsebenzi kanye nethuluzi eliqhubekayo lokuqashelwa ku-inthanethi (i-MemTrax) elisetshenziswe lapha futhi sibonga iziguli eziningi ezinokuwohloka komqondo ezifake isandla ocwaningweni oluyisisekelo olubalulekile. . Siphinde sibonge u-Xianbo Zhou kanye nozakwabo e-SJN Biomed LTD, ozakwabo kanye nabahlanganyeli ezibhedlela / ezindaweni zemitholampilo, ikakhulukazi uDkt. U-M. Luo noM. Zhong, abasize ngokuqashwa kwabahlanganyeli, ukuhlela izivivinyo, nokuqoqa, ukurekhoda, nokuphatha idatha, kanye nabahlanganyeli abangamavolontiya abanikele ngesikhathi sabo esibalulekile futhi bazibophezela ekuthatheni izivivinyo nokuhlinzeka. idatha enenani okufanele siyihlole kulolu cwaningo. Lokhu Ucwaningo lwasekelwa ngokwengxenye yi-MD Scientific Research Uhlelo lweNyuvesi Yezokwelapha yase-Kunming (Inombolo yesibonelelo sika-2017BS028 kuya ku-XL) kanye Nohlelo Lokucwaninga Lomnyango Wesayensi Nobuchwepheshe we-Yunnan (Umnikelo ongunombolo. 2019FE001 (-222) kuya ku-XL).

U-J. Wesson Ashford ufake isicelo selungelo lobunikazi ukuze kusetshenziswe ipharadigm ethile eqhubekayo yokuqashelwa echazwe kuleli phepha ngokujwayelekile. ukuhlolwa kwenkumbulo.

I-MemTrax, LLC yinkampani kaCurtis Ashford, futhi le nkampani iphethe i- ukuhlolwa kwenkumbulo uhlelo oluchazwe kuleli phepha.

Ukudalulwa kwababhali kuyatholakala ku-inthanethi (https://www.j-alz.com/manuscript-disclosures/19-1340r2).

inkumbulo test ukuwohloka komqondo ukuhlola inkumbulo ukulahlekelwa isivivinyo inkumbulo ukulahlekelwa ukuhlolwa inqama ukuhlola ingqondo ukudla okuhlukahlukene izincwadi ukuhlolwa kwengqondo ku-inthanethi
UCurtis Ashford - Umxhumanisi Wokucwaninga Ngengqondo

IZINDAWO ZOLWAZI

[1] I-Alzheimer's Association (2016) 2016 Alzheimer's amaqiniso kanye nezibalo. I-Alzheimers Dement 12, 459-509.
[2] Gresenz CR , Mitchell JM , Marrone J , Federoff HJ (2019) Umphumela wesigaba sokuqala Isifo i-Alzheimer emiphumeleni yezimali yasekhaya. I-Health Econ 29, 18–29.
[3] Foster NL , Bondi MW , Das R , Foss M , Hershey LA , Koh S , Logan R , Poole C , Shega JW , Sood A , Thothala N , Wicklund M , Yu M , Bennett A , Wang D (2019) Quality improvement in i-neurology: Isethi yokulinganisa ikhwalithi yokonakala kwengqondo okumaphakathi. I-Neurology 93, 705-713.
[4] Tong T , Thokala P , McMillan B , Ghosh R , Brazier J (2017) Ukusebenza kwezindleko zokusebenzisa izivivinyo zokuhlolwa kwengqondo zokuthola ukuwohloka komqondo kanye nokukhubazeka okuncane kokuqonda ekunakekelweni okuyisisekelo. I-Int J Geriatr Psychiatry 32, 1392-1400.
[5] Ashford JW , Gere E , Bayley PJ (2011) Ukulinganisa inkumbulo kuzilungiselelo zeqembu elikhulu kusetshenziswa ukuhlolwa kokuqashelwa okuqhubekayo. J Alzheimers Dis 27, 885-895.
[6] I-Ashford JW , Tarpin-Bernard F , Ashford CB , Ashford MT (2019) Umsebenzi wokuqashelwa oqhubekayo wekhompyutha wokulinganisa inkumbulo yesiqephu. J Alzheimers Dis 69, 385-399.
[7] I-Bergeron MF , Landset S , Tarpin-Bernard F , Ashford CB , Khoshgoftaar TM , Ashford JW (2019) Ukusebenza kwe-Episodic-memory kumodeli yokufunda komshini yokubikezela ukuhlukaniswa kwesimo sempilo yengqondo. J Alzheimers Dis 70, 277-286.
[8] van der Hoek MD , Nieuwenhuizen A , Keijer J , Ashford JW (2019) The Ukuhlolwa kwe-MemTrax uma kuqhathaniswa nesilinganiso sokuhlola kwengqondo sase-Montreal sokukhinyabezeka kwengqondo okuncane. J Alzheimers Dis 67, 1045-1054.
[9] U-Falcone M , Yadav N , Poellabauer C , Flynn P (2013) Ukusebenzisa imisindo yonkamisa ehlukanisiwe ukuze kuhlukaniswe ukulimala kobuchopho obuncane obulimazayo. Ngo-2013 I-IEEE International Conference on Acoustics, Speech and Signal Processing, Vancouver, BC, pp. 7577–7581.
[10] U-Dabek F, u-Caban JJ (2015) Ukusebenzisa idatha enkulu ukuze kufanekisele ithuba lokuthuthukisa izimo zengqondo ngemva kokungqubuzana. I-Procedia Comput Sci 53, 265–273.
[11] Climent MT , Pardo J , Munoz-Almaraz FJ , Guerrero MD , Moreno L (2018) Isihlahla sesinqumo sokutholwa kusenesikhathi kokukhubazeka kwengqondo ngosokhemisi bomphakathi. I-Front Pharmacol 9, 1232.
[12] I-Nasreddine ZS , Phillips NA , Bedirian V , Charbonneau S , Whitehead V , Collin I , Cummings JL , Chertkow H (2005) I-Montreal Cognitive Assessment, MoCA: Ithuluzi lokuhlola elifushane lokukhubazeka kwengqondo okuncane. J Am Geriatr Soc 53, 695–699.
[13] Yu J , Li J , Huang X (2012) Inguqulo yaseBeijing yokuhlola kwengqondo yase-Montreal njengethuluzi lokuhlola elifushane lokukhubazeka kwengqondo okumaphakathi: Ucwaningo olusekelwe emphakathini. I-BMC Psychiatry 12, 156.
[14] U-Chen KL , Xu Y , Chu AQ , Ding D , Liang XN , Nasreddine ZS , Dong Q , Hong Z , Zhao QH , Guo QH (2016) Ukuqinisekiswa kwenguqulo yesiShayina ye-Montreal yokuhlola kwengqondo okuyisisekelo ukuze kuhlolwe ukonakala kwengqondo okuncane. J Am Geriatr Soc 64, e285–e290.
[15] Carson N , Leach L , Murphy KJ (2018) Ukuhlolwa kabusha kwe-Montreal Cognitive Assessment (MoCA) scores cutoff. I-Int J Geriatr Psychiatry 33, 379–388.
[16] I-American Psychiatric Association (2013) I-Task Force Diagnostic and statistical manual of mental disorders: DSM-5™, American Psychiatric Publishing, Inc., Washington, DC.
[17] I-Python. I-Python Software Foundation, http://www.python.org, Ifinyelelwe ngoNovemba 15, 2019.
[18] Iqembu le-R Core, R: Ulimi nendawo yekhompyutha yezibalo R Isisekelo se-Statistical Computing, e-Vienna, e-Austria. https://www.R-project.org/, 2018, Ifinyelelwe ngoNovemba 15, 2019.
[19] Benavoli A , Corani G , Demšar J , Zaffalon M (2017) Isikhathi soshintsho: Isifundo sokuqhathanisa abaklami bezigaba abaningi ngokuhlaziywa kwe-Bayesian. J Mach Funda Res 18, 1–36.
[20] Frank E , Hall MA , Witten IH (2016) The WEKA Workbench. Ku Ukumbiwa Kwedatha: Amathuluzi Okufunda Omshini Angokoqobo Namasu, Frank E, Hall MA, Witten IH, Pal CJ, ed. Morgan Kaufmann https://www.cs.waikato.ac.nz/ml/weka/Witten_et_al_2016_appendix.pdf
[21] I-Bergeron MF , Landset S , Maugans TA , Williams VB , Collins CL , Wasserman EB , Khoshgoftaar TM (2019) Ukufunda ngomshini ekumodeleni uphawu lokungqubuzana kwezemidlalo esikoleni esiphakeme. I-Med Sci Sports Exerc 51, 1362-1371.
[22] Van Hulse J , Khoshgoftaar TM , Napolitano A (2007) Imibono yokuhlola yokufunda kusuka kudatha engalingani. Ku Izinqubo zeNgqungquthela Yamazwe Ngamazwe Yama-24 Yokufunda Ngomshini, Corvalis, Oregon, USA, amakhasi 935-942.
[23] Ashford JW , Kolm P , Colliver JA , Bekian C , Hsu LN (1989) Ukuhlolwa kwesiguli se-Alzheimer kanye nesimo sengqondo esincane: Ukuhlaziywa kwejika le-Item.P. J Gerontol 44, 139–146.
[24] Ashford JW , Jarvik L (1985) Isifo i-Alzheimer's: Ingabe i-neuron plasticity ibeka phambili ekuwohlokeni kwe-axonal neurofibrillary? N Engl J Med 313, 388–389.
[25] Jack CR Jr , Therneau TM , Weigand SD , ​​Wiste HJ , Knopman DS , Vemuri P , Lowe VJ , Mielke MM , Roberts RO , Machulda MM , Graff-Radford J , Jones DT , Schwarz CG , Senjenter JL , Rocca WA , Petersen RC (2019) Ukuvama kwezinhlangano ze-Alzheimer spectrum ezichazwe ngokwebhayoloji uma ziqhathaniswa ngokomtholampilo zisebenzisa i-National Institute on Aging-Alzheimer's Ucwaningo Lwenhlangano uhlaka. JAMA Neurol 76, 1174-1183.
[26] Zhou X , Ashford JW (2019) Intuthuko kumathuluzi okuhlola Isifo i-Alzheimer. Ukuguga Med 2, 88–93.
[27] Weiner MW , Nosheny R , Camacho M , Truran-Sacrey D , Mackin RS , Flenniken D , Ulbricht A , Insel P , Finley S , Fockler J , Veitch D (2018) The Brain Health I-Registry: Inkundla esekwe ku-inthanethi yokuqasha, ukuhlola, nokuqapha isikhathi eside ababambiqhaza ezifundweni ze-neuroscience. I-Alzheimers Dement 14, 1063-1076.
[28] Ashford JW , Schmitt FA (2001) Ukumodela inkambo yesikhathi ye Ukuwohloka komqondo kwe-Alzheimer. Curr Psychiatry Rep 3, 20–28.
[29] Li X , Wang X , Su L , Hu X , Han Y (2019) Sino Longitudinal Study on Cognitive Decline (SILCODE): Iphrothokholi yocwaningo lokubuka lwama-Chinese longitudinal longitudinal ukuthuthukisa amamodeli wokubikezela ubungozi okuguqulwa kuya ekukhubazekeni kwengqondo okumaphakathi kubantu abanokuqonda okuphathekayo nqaba. I-BMJ Open 9, e028188.
[30] Tarnanas I , Tsolaki A , Wiederhold M , Wiederhold B , Tsolaki M (2015) Iminyaka emihlanu yokuhlukahluka kwe-biomarker yeminyaka emihlanu I-Alzheimer's dementia isibikezelo: Ingabe umsebenzi wezinsimbi oyinkimbinkimbi womaka wokuphila kwansuku zonke ungavala izikhala? I-Alzheimers Dement (Amst) 1, 521-532.
[31] McGurran H , Glenn JM , Madero EN , Bott NT (2019) Ukuvimbela nokwelashwa kwesifo i-Alzheimer's: Izindlela zebhayoloji zokuzivocavoca. J Alzheimers Dis 69, 311-338.
[32] Mendiola-Precoma J , Berumen LC , Padilla K , Garcia-Alcocer G (2016) Ukwelapha ukuvimbela nokwelashwa kwesifo i-Alzheimer's. I-Biomed Res Int 2016, 2589276.
[33] Lane CA , Barnes J , Nicholas JM , Sudre CH , Cash DM , Malone IB , Parker TD , Keshavan A , Buchanan SM , Keuss SE , James SN , Lu K , Murray-Smith H , Wong A , Gordon E , Coath W , Modat M , Thomas D , Richards M , Fox NC , Schott JM (2020) Izinhlangano phakathi kwengozi yemithambo phakathi kwabantu abadala kanye ne-pathology yobuchopho empilweni yakamuva: Ubufakazi obuvela eqenjini laseBrithani lokuzalwa. I-JAMA Neurol 77, 175-183.
[34] I-Seshadri S (2020) Ukuvimbela ukucabanga kokuwohloka komqondo okungaphezu kweminyaka yobudala kanye namabhokisi e-amyloid. I-JAMA Neurol 77, 160-161.
[35] Maillard P , Seshadri S , Beiser A , Himali JJ , Au R , Fletcher E , Carmichael O , Wolf PA , DeCarli C (2012) Imiphumela ye-systolic blood pressure on white-matter integrity in young people in the Framingham Heart Study: A cross -isifundo sesigaba. I-Lancet Neurol 11, 1039–1047.
[36] Fink HA , Linskens EJ , Silverman PC , McCarten JR , Hemmy LS , Ouellette JM , Greer NL , Wilt TJ , Butler M (2020) Ukunemba kokuhlolwa kwe-biomarker ye-neuropathologically defined Isifo se-Alzheimer kubantu abadala asebekhulile abane-dementia. U-Ann Intern Med 172, 669-677.

Amanxusa: [a] SIVOTEC Analytics, Boca Raton, FL, USA | [b] Umnyango Wezobunjiniyela Bekhompyutha Nogesi kanye Nesayensi Yekhompyutha, eFlorida Atlantic University, Boca Raton, FL, USA | [c] SJN Biomed LTD, Kunming, Yunnan, China | [d] Isikhungo se Ucwaningo lwe-Alzheimer's, Washington Institute of Clinical Research, Washington, DC, USA | [e] Umnyango Wezokwelapha Zokuvuselela, Isibhedlela Sokuqala Esihlanganisiwe sase-Kunming Medical University, Kunming, Yunnan, China | [f] Umnyango Wesayensi Yezinzwa, Isibhedlela Sabantu sase-Dehong, Dehong, Yunnan, China | [g] Umnyango Wesayensi Yezinzwa, Isibhedlela Esisebenzisanayo Sokuqala sase-Kunming Medical University, i-Wuhua District, Kunming, Province of Yunnan, China | [h] Isikhungo Sokufunda Sokugula Nokulimala Okuhlobene Nempi, VA Palo Alto Ukunakekela impilo System, Palo Alto, CA, USA | [i] Umnyango Wesayensi Yengqondo Nezokuziphatha, i-Stanford University School of Medicine, i-Palo Alto, CA, USA

Ukuxhumana: [*] Izincwadi ezithunyelwa ku: Michael F. Bergeron, PhD, FACSM, SIVOTEC Analytics, Boca Raton Innovation Campus, 4800 T-Rex Avenue, Suite 315, Boca Raton, FL 33431, USA. I-imeyili: mbergron@sivotecanalytics.com.; U-Xiaolei Liu, MD, uMnyango Wezezinzwa, Isibhedlela Esisebenzisanayo Sokuqala saseKunming Medical University, 295 Xichang Road, Wuhua District, Kunming, Yunnan Province 650032, China. I-imeyili: ring@vip.163.com.

Amagama angukhiye: Ukuguga, Isifo i-Alzheimer, ukuwohloka komqondo, ukuhlolwa kwabantu abaningi