Նրա գրասենյակներ : Ittai Dayan Holger R. Roth Վիքիպահեստում Aoxiao Zhong-ի մասին Ahmed Harouni-ի մասին Amilcare Հավատություն Anas Z. Բիբինդին Բրիտանիա Liu Anthony Beardsworth Կոնտա Bradford J. Wood Վիքիպահեստում Սանհա Սանհա Սանհա Հիմնական հոդված՝ Chih-Hung Wang Հիմնական հոդված՝ Chun-Nan C. K. Լի Peiying Ruan Բրիտանիա Բրիտանիա Wu Eddie Huang Felipe Campos Kitamura-ի մասին Գրֆին Lacey Gustavo César de Antonio Corradi Վիքիպահեստում Գինո Նինո Հիմնական հոդվածներ Hirofumi Obinata Հիմա REN Jason C. Crane Jesse Tetreault-ի մասին Jiahui Guan- ը John W. Garrett-ի մասին Joshua D. Kaggie Jung Gil Park-ը Քեյս Dreyer Հիմնական հոդված՝ Krishna Juluru Քրթֆոր Քրթֆոր Marcio Aloisio Bezerra Cavalcanti- ը Rockenbach- ը Marius George Linguraru Վիքիպահեստում Մեսոն A. Haider Բրիտանիա AbdelMaseeh Nicola Rieke-ի մասին Pablo F. Damasceno Pedro Mario Cruz e Silva Հիմնական հոդված՝ Wang Sheng Xu Shuichi Kawano-ի մասին Բրիտանիա Բրիտանիա Soo Young Park Thomas M. Գրիտ Բրիտանիայի գրքեր Watsamon Jantarabenjakul Բրիտանիա Wang Young Tak-ի խաղը Xiang Li-ի մասին Xihong Լին Young Joon Kwon-ի մասին Հիմնական հոդված՝ Quraini Ապրիլ Feng Andrew N. Priest Վիքիպահեստում Բրիտանիա Turkbey Բրիտանիա Glicksberg Bernardo Bizzo Հիմնական հոդված՝ Kim Seok Carlos Tor-Díez Վիքիպահեստում Հիմնական հոդված՝ Chia-Cheng Lee Chia-Jung Hsu Չինաստան Lin Chiu-Ling Lai Christopher P. Hess-ի մասին Կոլին Compas Բրիտանիա Bhatia Eric K. Oermann-ի մասին Evan Leibovitz- ը Հիմնական հոդված՝ Sasaki Hitoshi Մորին Հիմնական հոդված Տեսակներ Jae Ho Son Հիմնական հոդված՝ Keshava Murthy Բրիտանիա Հիմնական հոդված՝ Հիմնական հոդված՝ Հիմնական հոդված՝ Հիմնական հոդվածը Մայք Fralick Մին Kyu Kang Mohammad Adil Natalie Gangai-ի մասին Բրիտանիա Vateekul Pierre Elnajjar-ի մասին Սարի Hickman Բրիտանիա Majumdar Shelley L. McLeod Sheridan Ռիդ Stefan Gräf Բրիտանիա Harmon Tatsuya Kodama Հիմնական հոդվածներ Tony Mazzulli-ի մասին Հիմնական հոդված՝ Vitor Lima de Trabor Yothin Rakvongthai Yu Rim Lee-ի մասին Հիմնական էջ Fiona J. Գիլբարթ Mona G. Flores Բրիտանիա Li Նրա գրասենյակներ : Աղջիկ Dayan Holger R. Roth Վիքիպահեստում Aoxiao Zhong-ի մասին Ahmed Harouni-ի մասին Amilcare Հավատություն Anas Z. Բիբինդին Բրիտանիա Liu Anthony Beardsworth Կոնտա Bradford J. Wood Վիքիպահեստում Սանհա Սանհա Սանհա Հիմնական հոդված՝ Chih-Hung Wang Հիմնական հոդված՝ Chun-Nan C. K. Լի Խաղադրություն Ruan Բրիտանիա Բրիտանիա Wu Բրիտանիա Huang Felipe Campos Kitamura-ի մասին Գրֆին Lacey Gustavo César de Antonio Corradi Վիքիպահեստում Գինո Նինո Հիմնական հոդվածներ Hirofumi Obinata Վիքիպահեստում Հիմա REN Jason C. Քարեն Jesse Tetreault-ի մասին Jiahui Guan- ը John W. Garrett-ի մասին Joshua D. Kaggie Վիքիպահեստում Jung Gil Park-ը Քեյս Dreyer Հիմնական հոդված՝ Krishna Juluru Քրթֆոր Քրթֆոր Marcio Aloisio Bezerra Cavalcanti- ը Rockenbach- ը Marius George Linguraru Վիքիպահեստում Մեսոն A. Haider Բրիտանիա AbdelMaseeh Nicola Rieke-ի մասին Pablo F. Damasceno Վիքիպահեստում Pedro Mario Cruz եւ Silva Հիմնական հոդված՝ Wang Sheng XU Shuichi Kawano-ի մասին Բրիտանիա Բրիտանիա Soo Young Park-ի մասին Thomas M. Գրիտ Բրիտանիայի գրքեր Սննդի սննդի սննդը Բրիտանիա Wang Young Tak-ի խաղը Xiang Li-ի մասին Xihong Լին Young Joon Kwon-ի մասին Հիմնական հոդված՝ Quraini Ապրիլ Feng Andrew N. Priest Վիքիպահեստում Բրիտանիա Turkbey Բրիտանիա Glicksberg Bernardo Bizzo-ի մասին Հիմնական հոդված՝ Kim Seok Carlos Tor-Díez Վիքիպահեստում Հիմնական հոդված՝ Chia-Cheng Lee Հիմնական հոդված՝ Chia-Jung Hsu Չինաստան Lin Գիտակներ Christopher P. Hess-ի մասին Կոլին Compas Բրիտանիա Bhatia Eric K. Oermann-ի մասին Evan Leibovitz- ը Հիմնական հոդված՝ Sasaki Hitoshi Մորին Հիմնական հոդված Տեսակներ Jae Ho Son Հիմնական հոդված՝ Keshava Murthy Բրիտանիա Հիմնական հոդված՝ Հիմնական հոդված՝ Հիմնական հոդված՝ Հիմնական հոդվածը Մայք Fralick Մին Kyu Kang Բրիտանիա Adil Natalie Gangai-ի մասին Բրիտանիա Vateekul Pierre Elnajjar-ի մասին Սարի Hickman Բրիտանիա Majumdar Shelley L. McLeod Վիքիպահեստում Sheridan Ռիդ Հիմնական հոդված՝ Stefan Gräf Բրիտանիա Harmon Հիմնական հոդվածներ Հիմնական հոդվածներ Tony Mazzulli-ի մասին Հիմնական հոդված՝ Vitor Lima de Trabor Հիմնական հոդված՝ Yothin Rakvongthai Yu Rim Lee-ի մասին Հիմնական էջ Fiona J. Գիլբարթ Mona G. Flores Վիքիպահեստում Բրիտանիա Li Ապրիլ Արդյոք, մենք օգտագործում ենք տվյալները 20 համալսարանների ամբողջ աշխարհում, որպեսզի ուսուցել FL մոդել, որը կոչվում է EXAM (elektronical medical record (EMR) chest X-ray AI մոդել), որը նախընտրում է ապացույցների հետազոտական հիվանդների հետ COVID-19- ի հետազոտական ալյումինե պահանջները, օգտագործելով ապացույցների ներքեւում, լաբորատորային տվյալները եւ գոտի X-ray. EXAM- ը հասկանում է միջին տարածք (AUC) >0.92 համար նախընտրել արդյունքները 24 եւ 72 ժամվա ընթացքում, այն ժամանակից, երբ առաջին ներկայացման է սենյակը, եւ այն ապահովել է 16% բարելավումը միջին AUC- ի հետազոտական մանրամասներ, որոնք մատակարարվում են բոլոր Գլխավոր Քաղաքական, գիտական, բժշկական եւ տվյալների գիտական ընկերություններ միացվել են միասին COVID-19- ի թեմայական ծախսում, որը արագ է գնահատել արագ եւ անվտանգ թեմայական ինտեգրման (AI) նոր թեմայք, եւ կարող են օգնել տվյալների փոխանցման եւ մոդելների դասընթացների եւ փորձարկման համար, առանց հագուստային համագործակցությունների հարմարավետության եւ տվյալների սեփականության բախսերի: , Healthcare մատակարարներ, հետազոտողներ եւ արդյունաբերության են վերցնել իրենց կենտրոնանալը պատասխանել անջատվել եւ կարեւոր բուժական պահանջները ստեղծվել է ծախսում, հետ հագուստ արդյունքների: , , , , , , Քաղաքական փորձերի վերահսկողությունը արագեցվել է եւ հեշտեցվել է ազգային կարգավիճակների եւ միջազգային համագործակցության հոգին: , , Data analytics- ի եւ AI- ի դասընթացները միշտ փնտրել են բաց եւ համագործակցական դիզայններ, ինչպիսիք են open-source ծրագրային ապահովման, reproducible research, data repositories- ը եւ anonymized data sets- ի հասանելիությունը. , Արդյունաբերական եւ տեխնոլոգիական ընկերությունների հետ միասին, տվյալների փոխանցման համար անհրաժեշտ է անմիջապես գործել տվյալների համագործակցությունը, որը կարող է հզորել klinische եւ գիտական ընկերություններ, երբ պատասխանել է արագ փոխանցման եւ լայնորեն ծախսերի աշխարհում: Data sharing has ethical, regulatory and legal complexities that are underscored, and somewhat complicated, by the recent entry of large technology companies into the healthcare data world. , , . 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 Այս CDS մոդելը մշակվել է Mass General Brigham- ում (MGB) եւ ստուգվել է բազմաթիվ առողջության համակարգերի տվյալների վրա: CDS մոդելի ներսում են բորսային X-ray (CXR) նկարները, vital signs, demographic տվյալները եւ լաբորատորային արժեքները, որոնք ցույց են տալիս, որ առաջադեմ գրասենյակներում կարող են տպագրել COVID-19- ի հաճախորդների արդյունքները , , , CXR- ը ընտրվել է որպես imaging input- ը, քանի որ այն լայնորեն հասանելի է եւ սովորաբար նշվում է Richtlinien- ի, ինչպիսիք են ACR- ի առաջարկում: Fleischner ընկերությունը Ով է WHO National Thoracic ընկերություններ National Ministry of Health COVID մոդելներ եւ radiology ընկերություններ ամբողջ աշխարհում CDS մոդելի արտադրանքը էր, որը կոչվում է CORISK , որը հարմարվում է սխալային աջակցության պահանջներին եւ որը կարող է օգնել սեղմելու հաճախորդներին Frontline Clinicists- ի կողմից , , Հավատված է, որ բժշկական ծառայողները սիրում են մոդելներ, որոնք ստուգված են իրենց սեփական տվյալների վրա: Ավելի շատ AI մոդելներ, այդ թվում `CDS մոդելը, ուսուցված են եւ հավելված են `կոտ` տվյալների վրա, որոնք հաճախ չկան տարբերություն: , , որը կարող է ազդել overfitting- ի եւ ցածր generalizability- ի վրա: Սա կարող է մանրածախվել մի քանի կայքերից տարբեր տվյալների հետ դասընթացով, առանց տվյալների կենտրոնացման: Օգտագործելով մեթոդներ, ինչպիսիք են transfer learning , FL- ը մի մեթոդ է օգտագործվում AI- ի մոդելների ուսուցման համար տարբեր data sources- ում, առանց տվյալների փոխանցման կամ արտահայտման իրենց առաջին տեղում: Այնպես որ, FL- ը շատ ոլորտներում կիրառելի է, վերջին ժամանակը առաջարկվել է cross-institutional healthcare research- ի համար: . 18 19 20 21 22 23 24 25 26 27 28 29 30 27 31 32 33 34 35 36 Federated Learning- ը աջակցում է կենտրոնական ձեւավորված փորձերի արագ սկսման հետ, որը բարելավում է տվյալների վերահսկողությունը եւ algorithmic փոխանակման եւ ազդեցության գնահատման One approach to FL, որը կոչվում է client-server, ուղեցնում է «նշելված» մոդել այլ սերվերի («նշելներ») համար, որոնք կատարում են մասական ուսուցման գործառույթներ, այնպես էլ ուղեցնում է արդյունքները վերցնել, որպեսզի ներգրավվել են կենտրոնական («նշելված») սերվով: Սա կատարվում է որպես սերտիֆիկ գործընթացը, մինչեւ ուսուցման ավարտվել է: . 37 36 FL- ի համար տվյալների կառավարությունը տեղականորեն պահպանվում է, փափուկ է անվտանգության խնդիրները, քանի որ միայն մոդելային քանակներ կամ gradients հաղորդում են հաճախորդային կայքեր եւ federated server- ի միջեւ: , FL- ը արդեն ցույց է տալիս, որ առաջադեմ բժշկական imaging ծրագրերի համար , , , COVID-19-ի հետազոտության մասին , , Հիմնական օրինակ է SARS-COV-2- ի հետ բուժված հաճախորդների mortality prediction մոդել, որը օգտագործում է բուժական առանձնահատկությունները, բայց սահմանափակ է մոդուլների եւ քանակությամբ: . 38 39 40 41 42 43 8 44 45 46 Մենք կարծում ենք, որ CDS- ի մոդելը կարող է հաջողությամբ federated լինել, քանի որ այն օգտագործվում է տվյալների ինտեգրվածները, որոնք հարմարավետ են բուժում, եւ որոնք ոչինչ չգիտեն մեծ մասը գործիչ-կախված փորձարկման հետ, որը կարող է օգնել սերտիֆիկի հաճախորդներին (հարկե, բուժական զգացմունքներ, կամ գրառված սմարթոմսերը): Հետեւաբար, լաբորատորային արդյունքները, կյանքի նշաններ, imaging ուսումնասիրություն, եւ սովորաբար ծածկված demographic (հարկե, տարի) օգտագործվել են: Մենք այնպես էլ վերահսկել CDS- ի մոդելը տարբեր տվյալների հետ, օգտագործելով client-server FL մոդել, որը ստեղծել է նոր Global FL մոդել, որը Մեր հստակությունը էր, որ EXAM- ը ավելի լավ կլինի, քան տեղական մոդելներ, եւ ավելի լավ կասկածվի առողջության համակարգերի մեջ: Ապրանքներ Exam մոդել Architecture- ը EXAM մոդելը հիմնված է վերանայված CDS մոդելով: Ապրանքի (հարկե, «դիմնական ճշմարտություն») տպագրերը կասկածվել են քիչ 24 եւ 72 ժամվա ընթացքում սեղմվող սեղմվող սեղմվող սեղմվող սեղմվող սեղմվող սեղմվող սեղմվող սեղմվող սեղմվող սեղմվող սեղմվող սեղմվող սեղմվող սեղմվող սեղմվող սեղմվող սեղմվող սեղմվող սեղմվող սեղմվող սեղմվող սեղմվող սեղմվող սեղմվող սեղմվող սեղմվող սեղմվող սեղմվող սեղմվող սեղմվող սեղմվող սեղմվող սեղմվող սեղմվող սեղմվող սեղմվող սեղմվող սեղմվող սեղմվող ս . 27 1 Ապրանքի մետաղադրամները հաճախորդների համար տեղադրվել են 0, 0,25, 0,50 եւ 0,75 մետաղադրամների համար, այնպես որ հաճախորդը ստացել է ամենամեծ օդի բուժում: Ապրանքի մետաղադրամների մետաղադրամների կետաղադրամները ներառում են սենյակային օդի (RA), ցածր օդի բուժում (LFO), բարձր օդի բուժում (HFO)/non-invasive ventilation (NIV) կամ մեխանիկական բուժում (MV). Եթե հաճախորդը մահացել է նախընտրման մետաղադրամում, արդյունքների մետաղադրամը տեղադրվել է 1: Այսպիսով, յուրաքանչյուր դեպքի համար կետաղադրամները ներառում են երկու մետաղադրամները 0-1- ում, ինչպիսիք EMR գործառույթների համար օգտագործվել են միայն ED- ում տպագրված առաջին արժեքները, եւ տվյալների preprocessing- ը ներառում է deidentification- ում, սխալ արժեքների տպագրման եւ normalization- ում մինչեւ 0-median եւ unity variance- ում: CXR բաղադրիչների համար օգտագործվել է միայն ED- ում տպագրված առաջին արժեքը: Մոդելը այնպես էլ բաղադրում է EMR- ի եւ CXR- ի առանձնահատկությունների տեղեկատվությունը, օգտագործելով 34-րդ մակերեսի Convolutional Neural Network- ը (ResNet34) CXR- ի եւ Deep & Cross ցանցի առանձնահատկությունները արտադրելու համար, որոնք բաղադրվում են EMR- ի առանձնահատկությունների հետ: ) Մոդելային արտադրանքը է Risk Score- ը, որը կոչվում է EXAM Score- ը, որը 0-1- ի մակարդակում է, յուրաքանչյուր 24-ի եւ 72 ժամի նախընտրականի համար, որը համեմատվում է նախընտրված մանրամասների հետ: մոդելներ Federating Մոդել EXAM մոդելը դասընթացվել է 16,148 բաղադրիչների հետ, ինչը ոչ միայն առաջին FL մոդելների միջեւ է COVID-19- ի համար, այլեւ շատ մեծ եւ բազմաբարձր բաղադրիչների զարգացման նախագծի մեջ klinisch relevant AI- ում (տեսանյութ) Data միջեւ կայքները չեն համատեղելի են առաջ արտադրման եւ, տեսնում է իրական կյանքի բուժական ինտերնետիկի պայմաններում, մանրամասական համատեղելիությունը տվյալների տպագրության չի կատարվել գրասենյակների կողմից (Fig. Ավելի 1A եւ B 1c եւ D World Map, որը ցույց է տալիս 20 տարբեր հաճախորդների կայքեր, որոնք օգնում են EXAM ուսումնասիրության. , Ապրանքների քանակը, որոնք ներառում են յուրաքանչյուր համալսարանը կամ կայքը (լուսանկարներ, որոնք ներառում են ամենամեծ քանակը հարցերի քանակը) Տղամարդկանց X-ray- ի հզորության մատակարարումը յուրաքանչյուր հաճախորդի կայքում: , Տեղադրողների կյանքը յուրաքանչյուր հաճախորդի կայքում, որը ցույց է տալիս minimum եւ maximum ages (stars), average age (triangles) եւ standard deviation (horizontal bars). The number of samples of each client site is shown in Additional Table . a b c d 1 Հիմնական հոդված՝ «Այս տարի» (տեսանյութ) - Հիմնական հոդված՝ «Այս տարի» (տեսանյութ) «1 × 10–3, Wilcoxon- ի նշված գոտի փորձը) 16% (հարկվում է միջին AUC- ի հետ, երբ մոդել կատարվում է masing-masing տեղական փորձի սարքերի վրա: 0.795-ից 0.920, կամ 12.5 մակարդակ) (Տեսանյութ) ) Այն նաեւ արդյունք է 38% generalizability բարելավումը (հարկե defined by average AUC when running the model on all test sets: from 0.667 to 0.920, կամ 25.3 մակարդակներ) լավագույն ամբողջ աշխարհում մոդել է նախընտրում 24-րդ ժամի օքսենային բուժման, համեմատաբար մոդելների հետ, որոնք ուսանվել են միայն կայքի սեփական տվյալների (Fig. Պահպանման արդյունքների համար 72-րդ ժամի սննդի բուժման համար, լավագույն ամբողջական մոդելային ուսուցման արդյունքների հետազոտական կատարման բարելավել է 18% - ի հետ, քան տեղականորեն ուսուցված մոդելների հետ, իսկ ամբողջական մոդելային բուժման բարելավել է միջին 34% - ի հետ (Extended Data Fig. ) Մեր արդյունքների հզորությունը ստուգվել է տեղական եւ FL դասընթացների երեք դասընթացների վերականգնման միջոցով տարբեր randomized տվյալների splits- ում: P 2A 2B 1 , Ապրանքի կատարումը յուրաքանչյուր հաճախորդի փորձարկման տեղադրված է 24-րդ ժամային սննդի վերամշակման նախընտրման համար մոդելների համար, որոնք ուսուցվել են միայն տեղական տվյալների վրա (Local) եւ լավագույն ամբողջական մոդելների համար, որոնք հասանելի են սերվերի վրա (FL)): Av., միջին փորձարկման կատարումը բոլոր կայքերում: , Generalizability (մոտական արդյունավետությունը այլ կայքերների փորձարկման տվյալների վրա, ինչպիսիք են միջին AUC) որպես գործառույթ է հաճախորդի տվյալների բաղադրիչի չափի (լուսանկարներ) (լուսանկարներ) ) եւ հաճախորդի 14- ից միայն RA- ի բուժման դեպքում, այնպես որ evaluation metric- ը (AUC- ի հետ) ոչինչ չի կիրառվում այդ դեպքում ( Data for client 14- ը նաեւ բացվել է տեղական մոդելների միջին generalizability- ի հաշվարկմանից: a b 1 մոդելներ Տեղական մոդելներ, որոնք դասընթացվել են օգտագործելով չորսակրված ախտորթները (հարկե, հիմնականում հեշտ բուժում COVID-19- ից), բացառապես օգտակար են FL- ից, քանի որ չափազանց բարելավված է հարմարեցման միջին AUC- ի կատարումը կատեգորիաների համար, որոնք ունեն միայն մի քանի բուժում: Սա բացահայտվել է հաճախորդի կայքում 16 (լուսանկարներ) (լուսանկարներ) (լուսանկարներ) (լուսանկարներ) (լուսանկարներ) (տեսանյութ) Բարձրացված տվյալների Fig. Ավելի կարեւոր է, որ FL մոդելի generalizability- ը շատ ավելի մեծ է, քան տեղական դասընթացային մոդելը: 3A 2 , ROC- ը client site- ում 16- ում, չժանգարեցված տվյալների հետ եւ հիմնականում հեշտ դեպքերով: Գլխավոր էջ » Գլխավոր էջ » Գլխավոր էջ » Գլխավոր էջ » Գլխավոր էջ » Գլխավոր էջ » Գլխավոր էջ » Գլխավոր էջ » Գլխավոր էջ » Գլխավոր էջ » Գլխավոր էջ » Գլխավոր էջ » Գլխավոր էջ » Գլխավոր էջ » Գլխավոր էջ » Գլխավոր էջ » Գլխավոր էջ » Գլխավոր էջ » Գլխավոր էջ » Գլխավոր էջ » Գլխավոր էջ » Գլխավոր էջ » Գլխավոր էջ » Գլխավոր էջ » Գլխավոր էջ » Գլխավոր էջ » Գլխավոր էջ » Գլխավոր էջ » Գլխավոր էջ » Գլխավոր էջ ) of the EXAM risk score are shown. Pos and neg denote the number of positive and negative cases, respectively, as defined by this range of EXAM score. a b t In the case of client sites with relatively small datasets, the best FL model markedly outperformed not only the local model but also those trained on larger datasets from five client sites in the Boston area of the USA (Fig. ). 3b The global model performed well in predicting oxygen needs at 24/72 h in patients both COVID positive and negative (Extended Data Fig. ). 3 Validation- ի անմիջական կայքերում Following initial training, EXAM was subsequently tested at three independent validation sites: Cooley Dickinson Hospital (CDH), Martha’s Vineyard Hospital (MVH) and Nantucket Cottage Hospital (NCH), all in Massachusetts, USA. The model was not retrained at these sites and it was used only for validation purposes. The cohort size and model inference results are summarized in Table , and the ROC curves and confusion matrices for the largest dataset (from CDH) are shown in Fig. Օպերացիոն կետը տեղադրվել է չորմատիկ օդանավման եւ մեխանիկական օդանավման (MV) բուժման (լուսանկարներ): FL- ի ամբողջ աշխարհում ուսուցված մոդելը, EXAM- ը, ստացել է 0.944- ի եւ 0.924- ի միջին AUC- ի համար 24- ի եւ 72-ի ժամի նախընտրական գործիքների համար: ), which exceeded the average performance among sites used in training EXAM. For prediction of MV treatment (or death) at 24 h, EXAM achieved a sensitivity of 0.950 and specificity of 0.882 at CDH, and a sensitivity of 1.000 specificity of 0.934 at MVH. NCH did not have any cases with MV/death at 24 h. In regard to 72-h MV prediction, EXAM achieved a sensitivity of 0.929 and specificity of 0.880 at CDH, sensitivity of 1.000 and specificity of 0.976 at MVH and sensitivity of 1.000 and specificity of 0.929 at NCH. 2 4 2 , , Performance (ROC) (կորեւ) եւ confusion matrices (կորեւ) EXAM FL մոդելի CDH տվյալների բաղադրիչի քաշի պահանջը 24 ժամում ( ) եւ 72 ժամ ( ) ROC- ի համար երեք տարբեր կատեգորիաների արժեքները ( ) վտանգավորության ստուգման սերտիֆիկը ցույց է տալիս. a b a b t CDH- ում 72- րոպեում, EXAM- ում 7.1%- ի ցածր սխալ-համփորդական քանակն էր: Բարձրացված տվյալների սխալները ներկայացվում են Extended Data Fig. , showing two false-negative cases from CDH where one case had many missing EMR data features and the other had a CXR with a motion artifact and some missing EMR features. 4 Հեղինակային տարբերակների օգտագործումը FL- ի օգտագործման հիմնական բաղադրիչն է իրենց տվյալների անվտանգությունը եւ անվտանգությունը, ինչպես նաեւ տվյալների համատեղելիությունը: FL-ի համար վերլուծման վտանգը պետք է լինի: or even the reconstruction of training images from the model gradients themselves . To counter these risks, security-enhancing measures were used to mitigate risk in the event of data ‘interception’ during site-server communication . We experimented with techniques to avoid interception of FL data, and added a security feature that we believe could encourage more institutions to use FL. We thus validated previous findings showing that partial weight sharing, and other differential privacy techniques, can successfully be applied in FL Արդյոք, այնպես էլ, մի մասային քաշի բաժանման համակարգը , , , մենք ցույց ենք տալիս, որ մոդելները կարող են հասնել հարմարավետ կատարումը նույնիսկ այն ժամանակ, երբ միայն 25% քանակի վերահսկողության համատեղված են (Extended Data Fig. ). 47 48 49 50 50 51 52 5 Discussion This study features a large, real-world healthcare FL study in terms of number of sites and number of data points used. We believe that it provides a powerful proof-of-concept of the feasibility of using FL for fast and collaborative development of needed AI models in healthcare. Our study involved multiple sites across four continents and under the oversight of different regulatory bodies, and thus holds the promise of being provided to different regulated markets in an expedited way. The global FL model, EXAM, proved to be more robust and achieved better results at individual sites than any model trained on only local data. We believe that consistent improvement was achieved owing to a larger, but also a more diverse, dataset, the use of data inputs that can be standardized and avoidance of clinical impressions/reported symptoms. These factors played an important part in increasing the benefits from this FL approach and its impact on performance, generalizability and, ultimately, the model’s usability. For a client site with a relatively small dataset, two typical approaches could be used for fitting a useful model: one is to train locally with its own data, the other is to apply a model trained on a larger dataset. For sites with small datasets, it would have been virtually impossible to build a performant deep learning model using only their local data. The finding, that these two approaches were outperformed on all three prediction tasks by the global FL model, indicates that the benefit for client sites with small datasets arising from participation in FL collaborations is substantial. This is probaby a reflection of FL’s ability to capture more diversity than local training, and to mitigate the bias present in models trained on a homogenous population. An under-represented population or age group in one hospital/region might be highly represented in another region—such as children who might be differentially affected by COVID-19, including disease manifestations in lung imaging . 46 Հիմնական արդյունքները հավատում են, որ ամբողջական մոդել հզոր է, որը աջակցում է մեր հավատումը, որ FL- ի ուսուցված մոդելներ կարող են հզորվել առողջապահական համակարգերի մեջ: Նրանք ապահովում են հավատալի գործառույթը COVID-19- ի հաճախորդների բուժում եւ FL- ի օգտագործման համար մոդել ստեղծման եւ փորձարկման մեջ: Այս ուսումնասիրության ներառումով հաճախորդների կայքները ստացել են EXAM- ի հասանելիությունը, որը պետք է ավելին հավատալվել, նախընտրելով ցանկացած վերահսկողության հավաստագրություն կամ կլանրաձայնային բուժում: Մեկտրոնային մոդելները պատրաստվում են հավատալել COVID-19- ի նպատակային ռեսուրսների օգտագործման համար MGB- ում `պրանքի , ինչպես նաեւ տարբեր կայքերում, որոնք ոչ մի մասը EXAM ուսուցման. 53 Over 200 prediction models to support decision-making in patients with COVID-19 have been published . Unlike the majority of publications focused on diagnosis of COVID-19 or prediction of mortality, we predicted oxygen requirements that have implications for patient management. We also used cases with unknown SARS-COV-2 status, and so the model could provide input to the physician ahead of receiving a result for PCR with reverse transcription (RT–PCR), making it useful for a real-life clinical setting. The model’s imaging input is used in common practice, in contrast with models that use chest computed tomography, a nonconsensual diagnostic modality. The model’s design was constrained to objective predictors, unlike many published studies that leveraged subjective clinical impressions. The data collected reflect varied incidence rates, and thus the ‘population momentum’ we encountered is more diverse. This implies that the algorithm can be useful in populations with different incidence rates. 19 Patient cohort identification and data harmonization are not novel issues in research and data science , բայց ավելի հարմարավետ են, երբ օգտագործում են FL- ը, քանի որ այլ կայքերում տվյալների սերտիֆիկները չեն տեսանելի: Կարդալիք են բուժական տեղեկատվության համակարգերի բարելավումներ, որոնք անհրաժեշտ են տվյալների պատրաստման հզորելու համար, որը առաջանում է FL- ում ներգրավված կայքի ցանցի ավելի լավ հարմարավետությունը: Սա, միասին hyperparameter- ի տեխնոլոգիաների հետ, կարող է տալիս ալբորմատոմսերը ավելի արդյունավետ «նշել» ավելի մեծ տվյալների սերտիֆիկների հետ, եւ մոդել parametrs- ից հարմարեցնել մի մասնավոր կայքում, որպեսզի ավելին հարմարեցնել, ինչպիսիք են, այդ կայքում ավելին ճնշման միջոցով: Մի համակարգ, որը թույլ է տալիս անմիջապես, ճշգրիտ ժամանակի մոդելային դիզայնի եւ արդյունքների մատակարարման, նաեւ կունենա, եւ կարող է «պահպանել դիզայնը» դասընթացից մոդելային տեղադրման համար: 54 39 Because data were not centralized they are not readily accessible. Given that, any future analysis of the results, beyond what was derived and collected, is limited. EXAM-ը, ինչպիսիք են այլ մեքենայի ուսուցման մոդելներ, ընդգրկված է դասընթացների տվյալների որակի վրա: Այս ալբորմատոմսը բուժման համար հետաքրքրված գործարանները պետք է գիտեն, որ դասընթացում կարող են հասկանալ պակասները: Օրինակ, EXAM-ի մոդելի դասընթացում օգտագործված նամակները ստացել են 24- ի եւ 72- ի ժամի քաշը: Հիմնվում է, որ քաշը, որը մատակարարվում է քաշի համար: Սակայն, COVID-19- ի պակասության առաջին փուլում, շատ քաշների համար բարձր քաշի քաշը մատակարարվել է պակասային քաշի համար, ոչ թե իրենց քաշի քաշի քաշը: Այս բուժում կարող է Քանի որ մեր տվյալների հասանելիությունը սահմանափակ էր, մենք չենք կարող ստանալ հարմարավետ տեղեկություններ, որոնք կարող են արտադրել ճշգրտված մանրամասներ, ինչպիսիք են սխալների պատճառները, post hoc- ում: Սակայն, մենք ուսումնասիրել ենք սխալները, որոնք գտնվում են ամենամեծ անմիջական փորձարկման կայքում, CDH- ում, եւ կարող ենք արտադրել սխալները, որոնք կարող ենք ստուգել ապագայում: Բարձր արդյունավետության կայքերում, կարծես, որ ամենամեծ սխալները ներառում են մեկի երկու կատեգորիաների մեջ: (1) ցածր որակի մանրամասների տվյալները, ինչպիսիք են CXR- ում սխալ տվյալները, կամ (2) արտադրանքի տվյալները, ինչպիսիք In future, we also intend to investigate the potential for a ‘population drift’ due to different phases of disease progression. We believe that, owing to the diversity across the 20 sites, this risk may have been mitigated. Մասնավորությունը, որը բարելավել է այդ տեսակի մեծ քանակի համագործակցությունը, այն է, որ կարող է նախընտրել յուրաքանչյուր հաճախորդի կայքի գործառույթը ամբողջական FL մոդելը բարելավելու համար: Սա օգնում է հաճախորդի կայքի ընտրությունը, եւ տվյալների ստուգման եւ ախտորատման գործառույթների նախընտրում: Հաջորդը հատկապես կարեւոր է, քանի որ այդ մեծ խոստոսի գործառույթների բարձր ծախսերը եւ դժվար ռեժիմիական գործառույթները, եւ դա թույլ է տալիս այդ գործառույթներին մուտքագրել տարբերություն, այլեւ տվյալների մանրամասների գեղեցիկ քանակը: Հաջորդ հոդված՝ Հաջորդ հոդված՝ Հաջորդ հոդված՝ Հաջորդ հոդված՝ Հաջորդ հոդված՝ Հաջորդ հոդված՝ Հաջորդ հոդված՝ Հաջորդ հոդված՝ Հաջորդ հոդված: Հաջորդ հոդված՝ Հաջորդ հոդված՝ Հաջորդ հոդված: Հաջորդ հոդված՝ Հաջորդ հոդված: Հաջորդ հոդված Neural Architecture Վիքիպահեստում and other automated machine learning Ապրանքներ, որոնք ավելի արդյունավետ են գտնել optimum training parameters համար յուրաքանչյուր հաճախորդի կայքը. 55 56 57 Գիտված խնդիրները բաղադրատոմսային normalization (BN) in FL Հիմնադրել ենք, որ մեր բազան մոդելը բեռնել է Image Feature Extraction- ի համար to reduce the divergence between unbalanced client sites. Future work might explore different types of normalization techniques to allow the training of AI models in FL more effectively when client data are nonindependent and identically distributed. 58 49 Լավագույն աշխատանքները, որոնք հիմնված են FL- ի տեղադրման մեջ պաշտպանվածության սխալների մասին, համոզված են, որ մոդելային դասընթացների ընթացքում տվյալների վախենում են: Միեւնույն ժամանակ, պաշտպանական algorithms դեռ անսահմանափակ են եւ սահմանափակ են բազմաթիվ գործառույթների: Երբ differential privacy algorithms , , show good protection, they may weaken the model’s performance. Encryption algorithms, such as homomorphic encryption Արդյոք, մի մանրամասն մեթոդ, որը կարող է չափազանց ավելացնել հաղորդագրության չափը եւ ուսուցման ժամանակը, թույլ է տալիս ավելի լավ ընտրություններ ստանալ, որոնք անհրաժեշտ են մանրամասների մանրամասների ստուգման համար, իսկ մանրամասների ստուգման ժամանակ մանրամասները կարող են մանրամասավորվել: , , . 59 36 48 49 60 36 48 49 Հաջորդ հավելվածության հետո, մենք տեսնում ենք, որ EXAM- ի մոդելը ED- ի կարգավիճակում տեղադրվում է, քանի որ այն կարող է գնահատել վտանգը, քանի որ ցանկացած հաճախորդի եւ տարածաշրջանի մակարդակում, եւ կօգնեն բուժողներին լրացուցիչ հարմարեցնել, երբ նրանք փորձում են փորձարկել հաճախորդներին: Մենք նաեւ տեսնում ենք, որ մոդելը օգտագործվում է ավելի հարմարավետ բուժման մակարդակային մետրոպիկով, որը օգնում է հարմարեցնել տարածքները, բուժողները եւ տարածքները: Մենք հույս ունենք, որ նույնպես FL- ի գործառույթները կարող են բեռնել տվյալների սիլոնները եւ թույլ տալ, որ շատ պահանջված AI- ի մոդելները ավելի արագ է զարգացման: Methods Ethics-ի հավելվածություն Բրիտանիայի ինտեգրված Բրիտանիայի ինտեգրված Բրիտանիայի ինտեգրված Բրիտանիայի ինտեգրված Բրիտանիայի ինտեգրված Բրիտանիայի ինտեգրված Բրիտանիայի ինտեգրված Բրիտանիայի ինտեգրված Բրիտանիայի ինտեգրված Բրիտանիայի ինտեգրված Բրիտանիայի ինտեգրված Բրիտանիայի ինտեգրված Բրիտանիայի ինտեգրված Բրիտանիայի ինտեգրված Բրիտանիայի ինտեգրված Բրիտանիայի ինտեգրված Բրիտանիայի ինտեգրված Բրիտանիայի ինտեգրված Բրիտանիայի ինտեգրված Բրիտանիայի ինտե MI-CLAIM guidelines for reporting of clinical AI models were followed (Supplementary Note ) 2 Study setting The study included data from 20 institutions (Fig. ): MGB, MGH, Brigham and Women's Hospital, Newton-Wellesley Hospital, North Shore Medical Center եւ Faulkner Hospital; Children's National Hospital in Washington, DC; NIHR Cambridge Biomedical Research Centre; The Self-Defense Forces Central Hospital in Tokyo; National Taiwan University MeDA Lab եւ MAHC եւ Taiwan National Health Insurance Administration; Tri-Service General Hospital in Taiwan; Kyungpook National University Hospital in South Korea; Faculty of Medicine, Chulalongkorn University in Thailand; Diagnosticos da America SA in Brazil; University of California, San Francisco; VA San Diego; University of Toronto; National Institutes of Health in Bethesda, Maryland; University of Wisconsin-Madison School of Medicine and Public Health; Memorial Sloan Kettering Cancer Center in New York; եւ Mount Sinai Health , , . Data from three independent sites were used for independent validation: CDH, MVH and NCH, all in Massachusetts, USA. These three hospitals had patient population characteristics different from the training sites. The data used for the algorithm validation consisted of patients admitted to the ED at these sites between March 2020 and February 2021, and that satisfied the same inclusion criteria of the data used to train the FL model. 1a 61 62 63 Data Collection-ի մասին The 20 client sites prepared a total of 16,148 cases (both positive and negative) for the purposes of training, validation and testing of the model (Fig. ). Medical data were accessed in relation to patients who satisfied the study inclusion criteria. Client sites strived to include all COVID-positive cases from the beginning of the pandemic in December 2019 and up to the time they started local training for the EXAM study. All local training had started by 30 September 2020. The sites also included other patients in the same period with negative RT–PCR test results. Since most of the sites had more SARS-COV-2-negative than -positive patients, we limited the number of negative patients included to, at most, 95% of the total cases at each client site. 1b A ‘case’ included a CXR and the requisite data inputs taken from the patient’s medical record. A breakdown of the cohort size of the dataset for each client site is shown in Fig. CXR տեսանյութի հզորության (մակելային արժեքների) թափանցման եւ մոդելների տարբերությունը մեծապես տարբեր էր կայքերում, քանի որ բազմաթիվ հաճախորդի եւ կայքի մասնավոր գործիչներ, ինչպիսիք են տարբեր սարքավորումների արտադրողները եւ տեսանյութի մոդելային մոդելները, ինչպիսիք են տեսանյութում: Արդյոք, ինչպիսիք են, որ EMR- ի գործառույթը, այնպես որ, ինչպիսիք են, որ EMR- ի գործառույթը, այնպես որ, ինչպիսիք են, որ EMR- ի գործառույթը, ինչպիսիք են, որ EMR- ի գործառույթը, այնպես որ, ինչպիսիք են, որ EMR- ի գործառույթը, այնպես որ, ինչպիսիք են, որ EMR- ի գործառույթը, այնպես որ, ինչպիսիք են, որ EMR- ի գործառույթը, ինչպիսիք են, որ EMR- ի գործառույթը, այնպես որ, ինչպիսիք են: Ավելի 1b 1c,d 6 Patient inclusion criteria Patient inclusion criteria were: (1) patient presented to the hospital’s ED or equivalent; (2) patient had a RT–PCR test performed at any time between presentation to the ED and discharge from the hospital; (3) patient had a CXR in the ED; and (4) patient’s record had at least five of the EMR values detailed in Table , all obtained in the ED, and the relevant outcomes captured during hospitalization. Of note, The CXR, laboratory results and vitals used were the first available for capture during the visit to the ED. The model did not incorporate any CXR, laboratory results or vitals acquired after leaving the ED. 1 Model input In total, 21 EMR features were used as input to the model. The outcome (that is, ground truth) labels were assigned based on patient requirements after 24- and 72-h periods from initial admission to the ED. A detailed list of the requested EMR features and outcomes can be seen in Table . 1 The distribution of oxygen treatment using different devices at different client sites is shown in Extended Data Fig. , which details the device usage at admission to the ED and after 24- and 72-h periods. The difference in dataset distribution between the largest and smallest client sites can be seen in Extended Data Fig. . 7 8 COVID-19- ի բաղադրիչների բաղադրիչը, որը հաստատվում է մեկ RT-PCR- ի ստուգում, որը ստացել է ED- ի հետազոտության եւ բաղադրիչի բաղադրիչի բաղադրիչի բաղադրիչի բաղադրիչի բաղադրիչի բաղադրիչի բաղադրիչի բաղադրիչի հետ: . Each client site was asked to randomly split its dataset into three parts: 70% for training, 10% for validation and 20% for testing. For both 24- and 72-h outcome prediction models, random splits for each of the three repeated local and FL training and evaluation experiments were independently generated. 1 EXAM model development There is wide variation in the clinical course of patients who present to hospital with symptoms of COVID-19, with some experiencing rapid deterioration in respiratory function requiring different interventions to prevent or mitigate hypoxemia , . A critical decision made during the evaluation of a patient at the initial point of care, or in the ED, is whether the patient is likely to require more invasive or resource-limited countermeasures or interventions (such as MV or monoclonal antibodies), and should therefore receive a scarce but effective therapy, a therapy with a narrow risk–benefit ratio due to side effects or a higher level of care, such as admittance to the intensive care unit . In contrast, a patient who is at lower risk of requiring invasive oxygen therapy may be placed in a less intensive care setting such as a regular ward, or even released from the ED for continuing self-monitoring at home EXAM- ը մշակվել է, որպեսզի օգնում է այդ հիվանդների սերտիֆիկացման. 62 63 64 65 Of note, the model is not approved by any regulatory agency at this time and it should be used only for research purposes. Տեղադրություն Score EXAM-ը ուսուցվել է FL- ի օգտագործման համար: Այն արտադրում է Risk Score- ում (հարկվում է EXAM Score- ում), ինչպիսիք են CORISK- ի (Տեսանյութի մանրամասները մանրամասներ են. ) and can be used in the same way to triage patients. It corresponds to a patient’s oxygen support requirements within two windows—24 and 72 h—after initial presentation to the ED. Extended Data Fig. illustrates how CORISK and the EXAM score can be used for patient triage. 27 9a 9B Տղամարդկանց X-ray բաղադրիչները նախընտրված են, որպեսզի ընտրել են նախնական տեղադրման բաղադրիչները եւ բացահայտել են lateral view բաղադրիչները, եւ այնուհետեւ չափազանցվել են 224 × 224- ի resolutions- ում: Որպես ցույց է տալիս Extended Data Fig. , the model fuses information from both EMR and CXR features (based on a modified ResNet34 with spatial attention Տեղադրված է CheXpert data set- ում) Deep & Cross ցանցը . To converge these different data types, a 512-dimensional feature vector was extracted from each CXR image using a pretrained ResNet34, with spatial attention, then concatenated with the EMR features as the input for the Deep & Cross network. The final output was a continuous value in the range 0–1 for both 24- and 72-h predictions, corresponding to the labels described above, as shown in Extended Data Fig. . We used cross-entropy as the loss function and ‘Adam’ as the optimizer. The model was implemented in Tensorflow NVIDIA Clara Train SDK-ի օգտագործումը Հիմնական AUC-ը դասընթացների համար (≥LFO, ≥HFO/NIV կամ ≥MV) թվարկվել է եւ օգտագործվել է որպես վերջնական դասընթացային մետրիկը, որը ստանդարտվել է սեղմվելի միջին եւ միավոր տարբերություն: CXR մանրամասները նախընտրված են ընտրել ճշգրիտ սերտիֆիկները եւ բացահայտել են բեռական տեսակի մանրամասները, ապա մանրամասները կասկածվել են 224 × 224 (լուսանկարներ: REF. Ավելի 9A 66 67 68 9B 69 70 27 Տեղադրման եւ normalization MissForest Algorithm- ը was used to impute EMR features, based on the local training dataset. If an EMR feature was completely missing from a client site dataset, the mean value of that feature, calculated exclusively on data from MGB client sites, was used. Then, EMR features were rescaled to zero-mean and unit variance based on statistics calculated on data from the MGB client sites. 71 EMR-CXR- ի տվյալների միացումը, օգտագործելով Deep & Cross ցանցը EMR- ի եւ CXR- ի տվյալների առանձնահատկությունների ինտերնետների մոդելման համար, Deep & Cross ցանցային դիզայնի վրա օգտագործվել է Deep-Feature Scheme- ը: Բիներային եւ կատեգորիաների առանձնահատկությունները EMR- ի մանրամասների համար, ինչպես նաեւ CXR- ում 512-dimensional տեսանյութերի առանձնահատկությունները, փոխվել են իրական արժեքների մանրամասների մանրամասների մեջ մանրամասների մանրամասների միջոցով: Բիներային եւ կատեգորիաների մանրամասները աշխատել են մանրամասների մանրամասների մանրամասների մանրամասների համար, որոնք մասնավորապես օգտագործել են մանրամասների մանրամասների մանրամասների միջեւ մանրամասների մանրամասների համար: Մանրամասների մանրամասների մանրամասները կատարել են մանրամասների մանրամասների մանրամասների մանրամասները իր մանրամաս 68 FL details Խնդրում ենք, որ FL- ի ամենամեծ հավելված ձեւը է Federated Average Algorithm- ի գործառույթը, ինչը McMahan et al- ը առաջարկում է: Այս դիզայնը կարող է կատարվել, օգտագործելով client-server տեղադրման, որտեղ յուրաքանչյուր մասնագիտացած կայքը աշխատում է որպես հաճախորդի: One can think of FL as a method aiming at minimizing a global loss function by reducing a set of local loss functions, which are estimated at each site. By minimizing each client site’s local loss while also synchronizing the learned client site weights on a centralized aggregation server, one can minimize global loss without having to access the entire dataset in a centralized location. Each client site learns locally, and shares model weight updates with a central server that aggregates contributions using secure sockets layer encryption and communication protocols. The server then sends an updated set of weights to each client site after aggregation sites, and resumes training locally. The server and Ավելի 72 9c A pseudoalgorithm of FL is shown in Supplementary Note Մեր փորձում, մենք տեղադրում ենք federated runs է = 200, մեկ տեղական դասընթացային շաբաթ մեկ շաբաթվա ընթացքում Որոշ հաճախորդների համար, այնքան շատ հաճախորդների համար, , մինչեւ 20 կախված է հաճախորդների ցանցային կապի կամ հասանելի տվյալների համար հատուկ նպատակային արդյունքների ժամանակի համար (24 կամ 72 ժամ). , հիմնված է յուրաքանչյուր հաճախորդի տվյալների չափի վրա Տեսանյութը օգտագործվում է յուրաքանչյուր հաճախորդի գործառույթների քանակման համար, երբ մոդելային քանակները միավորվում են federated averages- ում: FL- ի դասընթացների ժամանակ, յուրաքանչյուր հաճախորդի կայքը ընտրում է իր լավագույն տեղական մոդելը, հետազոտելով մոդելի կատարումը իր տեղական հավելվածության սարքում: Այսպիսով, սերվորը որոշում է լավագույն ամբողջական մոդել, որը հիմնված է յուրաքանչյուր հաճախորդի կայքում ուղարկված ամենամեծ հավելվածության քանակությամբ սերվորին: FL- ի դասընթացների հետո, լավագույն տեղական մոդելները եւ լավագույն ամբողջական մոդելը ավտոմատորեն զանգված են բոլոր հաճախորդի կայքերով եւ հավելված են իրենց տեղական ստուգման տվյալների վրա: 1 T t K Հիմնական k When training on local data only (the baseline), we set the epoch number to 200. The Adam optimizer was used for both local training and FL with an initial learning rate of 5 × 10–5 and a stepwise learning rate decay with a factor 0.5 after every 40 epochs, which is important for the convergence of federated averaging . Random affine transformations, including rotation, translations, shear, scaling and random intensity noise and shifts, were applied to the images for data augmentation during training. 73 Շնորհակալություն BN- ի բաղադրիչների Երբ աշխատում ենք տարբեր հաճախորդների հետ, ովքեր չեն անմիջապես եւ միասին թարմացվել են, մենք գտնել ենք, որ լավագույն մոդելային կատարումը տեղի է ունենում, երբ փնտրված ResNet34-ը տեղական ուշադրություն կտա: FL- ի դասընթացում սեղմված պարամետրերը (հարկե, օգտագործելով սեղմվող մակարդակների համար սեղմվող մակարդակը): Deep & Cross ցանցը, որը համատեղում է Image Features- ի հետ EMR- ի առանձնահատկությունները, չի ներառում BN- ի մակարդակները եւ այնպես էլ BN- ի անսահմանափակության խնդիրներ չեն ազդում: 58 47 Այս հետազոտության ընթացքում մենք ուսումնասիրել ենք անձնական կյանքի պահպանման համակարգը, որը համատեղում է միայն մասական մոդելների նորարարություններներ սերվերի եւ հաճախորդների կայքերների միջեւ: Տեսակների նորարարություններները կասկածվել են յուրաքանչյուր նորարարության ընթացքում բաղադրույքի չափով, եւ միայն մի որոշ մասը մեծագույն քանակների նորարարություններների հետ կասկածվել է սերվերի հետ: Հասկածապես, Տեսակների նորարարություններները (լուսանկարներ: gradients) կասկածվել են միայն այն ժամանակ, երբ իրենց անսահմանափակ արժեքը կասկած էր մի որոշ քանակը, (տեսանյութ) Ապրիշային տվյալների թարմացումը։ ), որը բաղադրված է բոլոր չորս բաղադրիչների, Δ , եւ կարող է տարբեր լինել յուրաքանչյուր հաճախորդի համար Մինչեւ ամեն թռիչք Այս scheme- ի տարբերակները կարող են ներառում մեծ gradients- ի լրացուցիչ կտրման կամ տարբերակային privacy schemes- ից: Նրանք ավելացնել են անմիջական խոշորությունը gradients- ում, կամ նույնիսկ Raw- ի տվյալների համար, նախքան սննդում են ցանցին: . k 5 Գլխավոր(t) k t 49 51 Statistical analysis We conducted a Wilcoxon signed-rank test to confirm the significance of the observed improvement in performance between the locally trained model and the FL model for the 24- and 72-h time points (Fig. Բարձրացված տվյալների Fig. Null hypothesis- ը բացահայտվել է միասին: «1 × 10–3 երկու դեպքում. 2 1 P Pearson- ի correlation- ը օգտագործվել է տեղական դասընթացների տեղական դասընթացների մոդելների համատեղելիությունը (մասնավոր AUC արժեքը այլ հաճախորդների կայքերների փորձարկման տվյալների հետ) համեմատելու համար. = 0,43 թ, = 0.035, ազատության մակարդակները (df) = 17 24-h մոդել եւ 0 0 62 = 0.003, df = 16 72-h մոդելի համար) Սա ցույց է տալիս, որ տվյալների բաղադրիչների չափը միայն ոչ մի գործիչ է, որը որոշում է մոդելի հզորությունը նստած տվյալների համար: r P r P ROC curves- ի համեմատումը Global FL մոդելից եւ տեղական մոդելների հետ, որոնք ուսուցվում են տարբեր կայքերում (Extended Data Fig. ), we bootstrapped 1,000 samples from the data and computed the resulting AUCs. We then calculated the difference between the two series and standardized using the formula = (AUC1 եւ AUC2) Ո՞վ is the standardized difference, is the standard deviation of the bootstrap differences and AUC1 and AUC2 are the corresponding bootstrapped AUC series. By comparing Երբ մենք ստանում ենք բուժել, մենք ստանում ենք Ապրանքներ, որոնք ներառում են Supplementary Table Ապրանքները ցույց են տալիս, որ null hypothesis- ը բացահայտվել է շատ ցածր չափով: Արդյոք, ինչպիսիք են այն, ինչպիսիք են այն, ինչպիսիք են այն, ինչպիսիք են այն, ինչպիսիք են այն, ինչպիսիք են այն, ինչպիսիք են այն, ինչպիսիք են այն, ինչպիսիք են այն, ինչպիսիք են այն, ինչպիսիք են այն, ինչպիսիք են այն, ինչպիսիք են այն, ինչպիսիք են այն, ինչպիսիք են այն, ինչպիսիք են նրանք: Գերմանները R- ում կատարվել են pROC գրասենյակով . 3 D s D s D P 2 P P 74 Քանի որ մոդելը նախընտրում է սխալ արտադրանքի, 0- ից 1- ում հետազոտական ստանդարտը, մի սխալ ստանդարտի գնահատումը, ինչպիսիք են qqplot- ը, չի կարող լինել: Հետեւաբար, ստանդարտի ստանդարտի մանրամասների համար մենք կատկել ենք ստանդարտությունը (Extended Data Fig. Մենք կատարում ենք միմյանց վերլուծություն տարբերության (ANOVA) փորձերը, որ համեմատել են տեղական եւ FL մոդելների ստանդարտների միջեւ շաբաթական ճշգրիտության կատեգորիաների (RA, LFO, HFO, MV). -Statistics- ը, որը մանրամասում է, որ մանրամասների միջեւ տարբերությունը կասկածված է մանրամասների ներքին տարբերությունը եւ մանրամասների միջեւ թափանցման մակարդակը, օգտագործվել է մոդելների կասկածման համար: Մեր արդյունքները ցույց են տալիս, որ -Այս տարբեր տեղական կայքի արժեքները 245.7, 253.4, 342.3, 389.8 եւ 634.8 են, իսկ FL մոդելի արժեքը 843.5 է: -մատները նշանակում են, որ թիմերը ավելի միասնական են, մեր FL մոդելից մանրամասները բացառապես ցույց են տալիս, որ գծերը ավելի մեծ են ծախսում cztery գծի ճշգրիտության կատեգորիաների միջեւ: Բացի այդ, ANOVA- ի արժեքը FL մոդելում <2 × 10–16 է, որը ցույց է տալիս, որ FL- ի նախընտրական քանակները տարբեր նախընտրական դասընթացների միջեւ statistically significantly տարբեր են: 10 F F F P Տեղադրման համոզվածություն Բարձր տեղեկություններ հետազոտական նախագծման մասին հասանելի են Փոխանցվել է այս հոդվածին. Nature Research-ի հետազոտական տեղեկատվություն Data հասանելիություն Այս տվյալները օգտագործվել են դասընթացների համար յուրաքանչյուր տեղում, եւ չեն փոխվել որեւէ այլ դասընթացների հետ կամ federated server- ի հետ, եւ նրանք չեն բացառապես հասանելի: CAMCA- ի կողմից վերահսկվում են ինտեգրված վերահսկման կայքի տվյալները, եւ մուտքագրումը կարող է պահանջվել Q.L- ի հետ: CAMCA- ի որոշման վրա, MGB- ի հետազոտական կառավարման կողմից եւ MGB- ի IRB- ի եւ քաղաքականության վրա կարող է իրականացնել տվյալների փոխանակման վերահսկման եւ վերահսկման համար հետազոտական նպատակների համար: Code- ի հասանելիություն Այս դասընթացում օգտագործված բոլոր կոդը եւ ծրագրային ապահովման համար NGC- ում բացառապես հասանելի են: Եթե ցանկանում եք ստանալ, ստանալ, կամ ստեղծել պրոֆիլը, ապա ներբեռնեք ներբեռնված URL- ից մեկը: Նկարագրված մոդելները, տվյալների պատրաստման ճշգրտություններ, դասընթացային կոդը, մոդելային ստուգման ստուգումը, readme ֆայլը, տեղադրման ճշգրտությունը եւ լվացքի ֆայլերը բացառապես հասանելի են NVIDIA NGC- ում: Տեսակ : Federated Learning Software-ը հասանելի է Clara Train SDK- ի մասում: Օգտագործեք այս մոդելը «wget --content-disposition» բեռնել -Այսը clara_train_covid19_exam_ehr_xray_1.zip» 61 https://ngc.nvidia.com/catalog/models/nvidia:med:clara_train_covid19_exam_ehr_xray https://ngc.nvidia.com/catalog/containers/nvidia:clara-train-sdk https://api.ngc.nvidia.com/v2/models/nvidia/med/clara_train_covid19_exam_ehr_xray/versions/1/zip Տեսակներ Budd, J. et al. Digital technologies in the public-health response to COVID-19. , 1183–1192 (2020). Nat. Med. 26 Moorthy, V., Henao Restrepo, A. M., Preziosi, M.-P. & Swaminathan, S. Data sharing for novel coronavirus (COVID-19). , 150 (2020). Bull. World Health Organ. 98 Chen, Q., Allot, A. & Lu, Z. Keep up with the latest coronavirus research. , 193 (2020). Nature 579 Fabbri, F., Bhatia, A., Mayer, A., Schlotter, B. & Kaiser, J. BCG IT spend pulse: how COVID-19 is shifting tech priorities. (2020). https://www.bcg.com/publications/2020/how-covid-19-is-shifting-big-it-spend Candelon, F., Reichert, T., Duranton, S., di Carlo, R. C. & De Bondt, M. The rise of the AI-powered company in the postcrisis world. (2020). https://www.bcg.com/en-gb/publications/2020/business-applications-artificial-intelligence-post-covid Chao, H. et al. Integrative analysis for COVID-19 patient outcome prediction. , 101844 (2021). Med. Image Anal. 67 Zhu, X. et al. Joint prediction and time estimation of COVID-19 developing severe symptoms using chest CT scan. , 101824 (2021). Med. Image Anal. 67 Yang, D. et al. Federated semi-supervised learning for Covid region segmentation in chest ct using multi-national data from China, Italy, Japan. , 101992 (2021). Med. Image Anal. 70 Minaee, S., Kafieh, R., Sonka, M., Yazdani, S. & Jamalipour Soufi, G. Deep-COVID: predicting COVID-19 from chest X-ray images using deep transfer learning. , 101794 (2020). Med. Image Anal. 65 COVID-19 Studies from the World Health Organization Database. (2020). https://clinicaltrials.gov/ct2/who_table ACTIV. (2020). https://www.nih.gov/research-training/medical-research-initiatives/activ Coronavirus Treatment Acceleration Program (CTAP). US Food and Drug Administration (2020). https://www.fda.gov/drugs/coronavirus-covid-19-drugs/coronavirus-treatment-acceleration-program-ctap Gleeson, P., Davison, A. P., Silver, R. A. & Ascoli, G. A. A commitment to open source in neuroscience. , 964–965 (2017). Neuron 96 Piwowar, H. et al. The state of OA: a large-scale analysis of the prevalence and impact of open access articles. , e4375 (2018). PeerJ. 6 European Society of Radiology (ESR). What the radiologist should know about artificial intelligence – an ESR white paper. , 44 (2019). Insights Imaging 10 Pesapane, F., Codari, M. & Sardanelli, F. Artificial intelligence in medical imaging: threat or opportunity? Radiologists again at the forefront of innovation in medicine. , 35 (2018). Eur. Radiol. Exp. 2 Price, W. N. 2nd & Cohen, I. G. Privacy in the age of medical big data. , 37–43 (2019). Nat. Med. 25 Liang, W. et al. Development and validation of a clinical risk score to predict the occurrence of critical illness in hospitalized patients with COVID-19. , 1081–1089 (2020). JAMA Intern. Med. 180 Wynants, L. et al. Prediction models for diagnosis and prognosis of covid-19 infection: systematic review and critical appraisal. , m1328 (2020). Brit. Med. J. 369 Zhang, L. et al. D-dimer levels on admission to predict in-hospital mortality in patients with Covid-19. , 1324–1329 (2020). J. Thromb. Haemost. 18 Sands, K. E. et al. Patient characteristics and admitting vital signs associated with coronavirus disease 2019 (COVID-19)-related mortality among patients admitted with noncritical illness. (2020). https://doi.org/10.1017/ice.2020.461 American College of Radiology. CR recommendations for the use of chest radiography and computed tomography (CT) for suspected COVID-19 infection. (2020). https://www.acr.org/Advocacy-and-Economics/ACR-Position-Statements/Recommendations-for-Chest-Radiography-and-CT-for-Suspected-COVID19-Infection Rubin, G. D. et al. The role of chest imaging in patient management during the COVID-19 pandemic: a multinational consensus statement from the Fleischner Society. , 172–180 (2020). Radiology 296 World Health Organization. Use of chest imaging in COVID-19. (2020). https://www.who.int/publications/i/item/use-of-chest-imaging-in-covid-19 Jamil, S. et al. Diagnosis and management of COVID-19 disease. , 10 (2020). Am. J. Respir. Crit. Care Med. 201 Redmond, C. E., Nicolaou, S., Berger, F. H., Sheikh, A. M. & Patlas, M. N. Emergency radiology during the COVID-19 pandemic: The Canadian Association of Radiologists Recommendations for Practice. , 425–430 (2020). Can. Assoc. Radiologists J. 71 Buch, V. et al. Development and validation of a deep learning model for prediction of severe outcomes in suspected COVID-19 Infection. Preprint at (2021). https://arxiv.org/abs/2103.11269 Lyons, C. & Callaghan, M. The use of high-flow nasal oxygen in COVID-19. , 843–847 (2020). Anaesthesia 75 Whittle, J. S., Pavlov, I., Sacchetti, A. D., Atwood, C. & Rosenberg, M. S. Respiratory support for adult patients with COVID-19. , 95–101 (2020). J. Am. Coll. Emerg. Physicians Open 1 Ai, J., Li, Y., Zhou, X. & Zhang, W. COVID-19: treating and managing severe cases. , 370–371 (2020). Cell Res. 30 Esteva, A. et al. A guide to deep learning in healthcare. , 24–29 (2019). Nat. Med. 25 Cahan, E. M., Hernandez-Boussard, T., Thadaney-Israni, S. & Rubin, D. L. Putting the data before the algorithm in big data addressing personalized healthcare. , 78 (2019). NPJ Digit. Med. 2 Thrall, J. H. et al. Artificial intelligence and machine learning in radiology: opportunities, challenges, pitfalls, and criteria for success. , 504–508 (2018). J. Am. Coll. Radiol. 15 Shilo, S., Rossman, H. & Segal, E. Axes of a revolution: challenges and promises of big data in healthcare. , 29–38 (2020). Nat. Med. 26 Gao, Y. & Cui, Y. Deep transfer learning for reducing health care disparities arising from biomedical data inequality. , 5131 (2020). Nat. Commun. 11 Rieke, N. et al. The future of digital health with federated learning. , 119 (2020). NPJ Dig. Med. 3 Yang, Q., Liu, Y., Chen, T. & Tong, Y. Federated machine learning: concept and applications. , 12 (2019). ACM Trans. Intell. Syst. Technol. 10 Ma, C. et al. On safeguarding privacy and security in the framework of federated learning. , 242–248 (2020). IEEE Netw. 34 Brisimi, T. S. et al. Federated learning of predictive models from federated Electronic Health Records. , 59–67 (2018). Int. J. Med. Inform. 112 Roth, H. R. et al. Federated learning for breast density classification: a real-world implementation. In , (eds. Albarqouni, S. et al.) Vol. 12,444, 181–191 (Springer International Publishing, 2020). Proc. Second MICCAI Workshop, DART 2020 and First MICCAI Workshop, DCL 2020 Domain Adaptation and Representation Transfer, and Distributed and Collaborative Learning Sheller, M. J. et al. Federated learning in medicine: facilitating multi-institutional collaborations without sharing patient data. , 12598 (2020). Sci. Rep. 10 Remedios, S. W., Butman, J. A., Landman, B. A. & Pham, D. L. in (eds Remedios, S. W. et al.) (Springer, 2020). Federated Gradient Averaging for Multi-Site Training with Momentum-Based Optimizers Xu, Y. et al. A collaborative online AI engine for CT-based COVID-19 diagnosis. Preprint at (2020). https://www.medrxiv.org/content/10.1101/2020.05.10.20096073v2 Raisaro, J. L. et al. SCOR: A secure international informatics infrastructure to investigate COVID-19. , 1721–1726 (2020). J. Am. Med. Inform. Assoc. 27 Vaid, A. et al. Federated learning of electronic health records to improve mortality prediction in hospitalized patients with COVID-19: machine learning approach. , e24207 (2021). JMIR Med. Inform. 9 Nino, G. et al. Pediatric lung imaging features of COVID-19: a systematic review and meta-analysis. , 252–263 (2021). Pediatr. Pulmonol. 56 Fredrikson, M., Jha, S. & Ristenpart, T. Model inversion attacks that exploit confidence information and basic countermeasures. In 1322–1333, (2015). Proc. 22nd ACM SIGSAC Conference on Computer and Communications Security https://doi.org/10.1145/2810103.2813677 Zhu, L., Liu, Z. & Han, S. in (eds Wallach, H. et al.) 14774–14784 (Curran Associates, Inc., 2019). Advances in Neural Information Processing Systems 32 Kaissis, G. A., Makowski, M. R., Rückert, D. & Braren, R. F. Secure, privacy-preserving and federated machine learning in medical imaging. , 305–311 (2020). Nat. Mach. Intell. 2 Li, W. et al. in 133–141 (Springer, 2019). Privacy-Preserving Federated Brain Tumour Segmentation Shokri, R. & Shmatikov, V. Privacy-preserving deep learning. In (2015). Proc. 53rd Annual Allerton Conference on Communication, Control, and Computing (Allerton) https://doi.org/10.1109/allerton.2015.7447103 Li, X. et al. Multi-site fMRI analysis using privacy-preserving federated learning and domain adaptation: ABIDE results. , 101765 (2020). Med. Image Anal. 65 Estiri, H. et al. Predicting COVID-19 mortality with electronic medical records. , 15 (2021). NPJ Dig. Med. 4 Jiang, G. et al. Harmonization of detailed clinical models with clinical study data standards. , 65–74 (2015). Methods Inf. Med. 54 Yang, D. et al. in . (2019). Searching Learning Strategy with Reinforcement Learning for 3D Medical Image Segmentation https://doi.org/10.1007/978-3-030-32245-8_1 Elsken, T., Metzen, J. H. & Hutter, F. Neural architecture search: a survey. , 1–21 (2019). J. Mach. Learning Res. 20 Yao, Q. et al. Taking human out of learning applications: a survey on automated machine learning. Preprint at (2019). https://arxiv.org/abs/1810.13306 Ioffe, S. & Szegedy, C. Batch normalization: accelerating deep network training by reducing internal covariate shift. In , PMLR , 448–456 (2015). Proc. 32nd International Conf. Machine Learning 37 Kaufman, S., Rosset, S. & Perlich, C. Leakage in data mining: formulation, detection, and avoidance. In , 556–563 (2011). Proc. 17th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining Zhang, C. et al. BatchCrypt: efficient homomorphic encryption for cross-silo federated learning. In , 493–506 (2020). Proc. 2020 USENIX Annual Technical Conference, ATC 2020 . (2020). Nvidia NGC Catalog: COVID-19 Related Models https://ngc.nvidia.com/catalog/models?orderBy=scoreDESC&pageNumber=0&query=covid&quickFilter=models&filters Marini, J. J. & Gattinoni, L. Management of COVID-19 respiratory distress. , 2329–2330 (2020). JAMA 323 Cook, T. M. et al. Consensus guidelines for managing the airway in patients with COVID-19: Guidelines from the Difficult Airway Society, the Association of Anaesthetists the Intensive Care Society, the Faculty of Intensive Care Medicine and the Royal College of Anaesthetist. , 785–799 (2020). Anaesthesia 75 Galloway, J. B. et al. A clinical risk score to identify patients with COVID-19 at high risk of critical care admission or death: an observational cohort study. , 282–288 (2020). J. Infect. 81 Kilaru, A. S. et al. Return hospital admissions among 1419 COVID-19 patients discharged from five U.S. emergency departments. , 1039–1042 (2020). Acad. Emerg. Med. 27 He, K., Zhang, X., Ren, S. & Sun, J. Deep residual learning for image recognition. In (2016). Proc. 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) https://doi.org/10.1109/cvpr.2016.90 Irvin, J. et al. CheXpert: a large chest radiograph dataset with uncertainty labels and expert comparison. , 590–597 (2019). Proc. AAAI Conf. Artif. Intell. 33 Wang, R., Fu, B., Fu, G. & Wang, M. Deep & Cross network for Ad Click predictions. In Article no. 12 (2017). Proc. ADKDD’17 Abadi, M. et al. TensorFlow: asystem for large-scale machine learning. In , USENIX Association 265–283 (2016). 12th USENIX Symposium on Operating Systems Design and Implementation (OSDI 16) . (2020). NVIDIA Clara Imaging https://developer.nvidia.com/clara-medical-imaging Stekhoven, D. J. & Bühlmann, P. MissForest–non-parametric missing value imputation for mixed-type data. , 112–118 (2012). Bioinformatics 28 McMahan, H., Moore, E., Ramage, D., Hampson, S. & y Arcas, B. A. Communication-efficient learning of deep networks from decentralized data. (2017). http://proceedings.mlr.press/v54/mcmahan17a.html Hsieh, K., Phanishayee, A., Mutlu, O. & Gibbons, P. B. The non-IID data quagmire of decentralized machine learning. In PMLR 119 (2020). Proc. 37th International Conf. Machine Learning Robin, X. et al. pROC: an open-source package for R and S+ to analyze and compare ROC curves. , 77 (2011). BMC Bioinformatics 12 Հիմնական տեղեկություններ Տեսակները, որոնք հայտնաբերվել են այս հետազոտության մեջ, հավատում են, որ գրողները են, եւ ոչ միայն NHS- ի, NIHR- ի, Health and Social Care Department- ի կամ ցանկացած մասնագիտության միջոցով: MGB- ի հետազոտական տվյալները, որոնք հավատում են: J. Brink, Radiology Department, Massachusetts General Hospital, Harvard Medical School, Boston, MA; M. Kalra, Radiology Department, Massachusetts General Hospital, Harvard Medical School, Boston, MA; N. Neumark, Center for Clinical Data Science, Massachusetts General Brigham, Boston, MA; T. Schultz, Radiology Department, Massachusetts General Hospital, Boston, MA; N. Guo, Center for Advanced Medical Computing and Analysis, Study Department, Massachusetts General Hospital, Harvard Medical School, Boston, MA; J. K. Cram Շնորհակալություն, որ Մարդկային համալսարանը, Chulalongkorn համալսարանը, Շնորհակալություն, որ Ratchadapisek Sompoch Endowment Fund RA (PO) (No. 001/63)- ը COVID-19- ի հետ կապված բուժական տվյալների եւ բուժական նմուշների հավաքման եւ կառավարման համար: Մարդկային համալսարանը, Chulalongkorn համալսարան. NIHR Cambridge Biomedical Research Centre- ը Շնորհակալություն է NIHR- ի (Cambridge Biomedical Research Centre- ի) աջակցությամբ: National Taiwan University MeDA Lab- ը եւ MAHC- ը եւ Taiwan National Health Insurance Administration- ը: MOST Joint Research Center for AI technology- ը, All Vista Healthcare National Health Insurance Administration- ը, Taiwan- ը, Մարդկության եւ տեխնոլոգիաների https://data.ucsf.edu/covid19 Այս գրասենյակը հասանելի է CC by 4.0 Deed (Attribution 4.0 International) License- ի տակ: Այս թերթը CC by 4.0 Deed- ի (Attribution 4.0 International) լիցենզիաների կողմից: Արդյոք հասանելի է բնական