Kwenzekeni
Abacwaningi e-Carnegie Mellon University kanye ne-UPMC bathuthukise i-algorithm yokufunda yomshini engagadiwe ukuze babone i-vasculopathy ye-decidual ku-placental tissue, inkomba eyinhloko ye-postpartum preeclampsia. Ithuluzi lihlaziya inhlangano yendawo yamaseli e-trophoblast e-extravillous kanye namaseli abomvu egazi ukuze ahlabe umkhosi amacala adinga ukubuyekezwa okukhethekile, okuhloswe ngawo ukukhulisa amazinga okuhlolwa e-US lapho kuhlolwa khona angaphansi kuka-20% wama-placentas okwamanje.
Abacwaningi eMnyangweni Wezobunjiniyela Bemishini e-Carnegie Mellon University kanye noMnyango Wezifo Zomzimba e-UPMC bathuthukise i-algorithm yokufunda ngomshini ukuze kwenziwe lula ukuhlolwa kwe-placenta ye-preeclampsia. Lolu cwaningo, olushicilelwe ku-npj Digital Medicine, ludingida inselele yokuthi angaphansi kuka-20% wama-placenta e-United States ahlolwayo ngemva kokubeletha ngenxa yokushoda okukhulu kodokotela bezifo eziphathelene nokubeletha abaqeqeshiwe.
I-algorithm igxile ekuboneni i-decidual vasculopathy, isifo semithambo yegazi kamama ku-placenta ehambisana ne-postpartum preeclampsia. Imodeli isebenzisa izakhiwo zokudlulisa ukukhanya ezincike ku-stain ukuze ihlaziye inhlangano yendawo yamaseli e-extravillous trophoblast (EVT) namaseli abomvu egazi ngaphakathi kwemithambo ye-placental. Imikhumbi enempilo inama-EVT ambalwa ahlobene namangqamuzana egazi abomvu, kuyilapho imikhumbi egulayo inokuningi, okwenza amaphethini ahlukene i-AI engawathola.
Isici esibalulekile semodeli 'isikolo sokuhlukaniswa kwe-morphology,' esilinganisa ukuthi inhlangano yeseli ijwayelekile noma ayijwayelekile kangakanani. Le ndlela ivumela i-AI ukuthi ingagcini nje ngokuhlaba umkhosi isifo esingaba khona kodwa futhi ichaze isisekelo sebhayoloji sesiphetho saso, isici esibalulekile sokwakha ukwethenjwa ezinhlelweni ze-AI zomtholampilo. Abacwaningi bachaza le nqubo njengefana nokuthola imikhumbi ethile enesifo phakathi kwezinkulungwane ezibukeka zinempilo efanayo.
Imininingwane yomthombo: medicalxpress.com ↗
Kungani kubalulekile
I-Preeclampsia iyimbangela ehamba phambili yokufa okuhlobene nokukhulelwa, nokho amandla okuxilonga anqunyelwe kakhulu ukushoda kodokotela bezifo eziphathelene nokubeletha. Leli thuluzi le-AI linikeza isixazululo esinokwethenjelwa sokwenza ukuhlola kokuqala ngokuzenzakalela, okungase kuvumele izazi zezifo ukuthi zibuyekeze izimo eziningi futhi zithole izinkinga zangemva kokubeletha ngaphambi kwesikhathi. Ukugxila kwemodeli ekuchazekeni, kusetshenziswa 'isikolo sokuhlukaniswa kwe-morphology' ukuze kugqanyiswe umehluko wezinto eziphilayo, kusingathwa umgoqo omkhulu ekuthathweni komtholampilo ngokunikeza ukucabanga okusobala esikhundleni sokuphuma kwebhokisi elimnyama.
I-Preeclampsia iseyimbangela ehamba phambili yokufa okuhlobene nokukhulelwa, futhi izinkinga zangemva kokubeletha zingavela ezinsukwini noma amasonto ngemva kokubeletha. Ibhodlela lamanje lokuxilonga amandla omuntu; ngaphandle kokuhlolwa okuzenzakalelayo, iziguli eziningi ezisengozini zingase zingakutholi ukuhlonzwa okufika ngesikhathi kokungajwayelekile kwe-placenta.
Ukuthuthukiswa kwemodeli ye-AI echazekayo kubalulekile ekutholweni kokunakekelwa kwezempilo. Ngokuhlinzeka ngomehluko olinganisekayo webhayoloji kunokuxilongwa kwebhokisi elimnyama, ithuluzi lihlose ukwenza kube lula ukwethembana phakathi kwezazi zezifo. Lokhu kungafihli kubalulekile ekuhlanganiseni i-AI ezinqumweni zezokwelapha ezisezingeni eliphezulu lapho ukuziphendulela kubaluleke kakhulu.
Umthelela ongaba khona udlulela ngale kwe-preeclampsia. Abacwaningi basikisela ukuthi amamodeli e-AI asekelwe ku-biomarker afanayo angasetshenziswa kwezinye izimo, njengomdlavuza kanye nezifo ezihlobene nobuchopho, lapho ukutholakala kusenesikhathi kwezinguquko ezithile zamaselula kubalulekile emiphumeleni yokwelashwa.
I-Interactive Mechanism: Indlela Esebenza Ngayo Ngempela
Hlola ubuchwepheshe obuyisisekelo ngemuva kwalokhu kuthuthukiswa ngokuhlanganyela.
Why can ethical evaluation not be reduced to one model score?
Ongakubuka ngokulandelayo
Izifundo zokuqinisekisa umtholampilo ukuze kutholwe ukuzwela kwe-algorithm nokucaciswa kwezilungiselelo zesibhedlela somhlaba wangempela. Izindlela zokulawula zokugunyazwa kwe-FDA kwesoftware njengethuluzi lezokwelapha. Ukuhlanganiswa kwethuluzi ekuhambeni komsebenzi okukhona kwesibhedlela se-pathology kanye nomthelela walo esilinganisweni esiphelele sokuhlolwa kwe-placenta.
Isinyathelo esilandelayo esibalulekile siwukuqinisekisa okuzimele komtholampilo ukuze kuqinisekiswe ukunemba kwe-algorithm kubantu abahlukahlukene beziguli. Nakuba ucwaningo lubonisa ukuba nokwenzeka kobuchwepheshe, ukusebenza komhlaba wangempela kuzilungiselelo zesibhedlela kuzonquma ukusetshenziswa kwalo komtholampilo.
Kuzodingeka imvume yokulawula ngaphambi kokuba ithuluzi lisatshalaliswe kabanzi. Uhlaka lwe-FDA lwemishini yezokwelapha esekelwe ku-AI cishe luzodinga ukuhlolwa okuqinile ukuze kuqinisekiswe ukuphepha nokusebenza ngempumelelo, okungathatha iminyaka eminingana.
Izinhlelo zokunakekelwa kwezempilo zizodinga ukuhlola ukusebenza kahle kwezindleko zokusebenzisa amathuluzi okuhlola anjalo. Uma i-AI ingakhuphula kakhulu umthamo we-placenta ebuyekezwe odokotela bezifo, ingaholela ekungeneleleni kwangaphambili kanye nemiphumela yezempilo kamama ethuthukisiwe.