Federated Learning used for predicting outcomes in SARS-COV-2 patients.
Mona G FloresIttai DayanHolger R RothAoxiao ZhongAhmed HarouniAmilcare GentiliAnas AbidinAndrew LiuAnthony CostaBradford WoodChien-Sung TsaiChih-Hung WangChun-Nan HsuC K LeeColleen RuanDaguang XuDufan WuEddie HuangFelipe Campos KitamuraGriffin LaceyGustavo César de Antônio CorradiHao-Hsin ShinHirofumi ObinataHui RenJason CraneJesse TetreaultJiahui GuanJohn W GarrettJung Gil ParkKeith DreyerKrishna JuluruKristopher KerstenMarcio Aloisio Bezerra Cavalcanti RockenbachMarius LinguraruMasoom HaiderMeena AbdelMaseehNicola RiekePablo DamascenoPedro Mario Cruz E SilvaPochuan WangSheng XuShuichi KawanoSira SriswasdiSoo Young ParkThomas GristVarun BuchWatsamon JantarabenjakulWeichung WangWon Young TakXiang LiXihong LinFred KwonFiona J GilbertJosh KaggieQuanzheng LiAbood QurainiAndrew FengAndrew N PriestIsmail Baris TurkbeyBenjamin Scott GlicksbergBernardo Canedo BizzoByung Seok KimCarlos Tor-DiezChia-Cheng LeeChia-Jung HsuChin LinChiu-Ling LaiChristopher HessColin CompasDeepi BhatiaEric OermannEvan LeibovitzHisashi SasakiHitoshi MoriIsaac YangJae Ho SohnKrishna Nand Keshava MurthyLi-Chen FuMatheus Ribeiro Furtado de MendonçaMike FralickMin Kyu KangMohammad AdilNatalie GangaiPeerapon VateekulPierre ElnajjarSarah HickmanSharmila MajumdarShelley McLeodSheridan ReedStefan GräfStephanie A HarmonTatsuya KodamaThanyawee PuthanakitTony MazzulliVitor de Lima LavorYothin RakvongthaiYu Rim LeeYuhong WenPublished in: Research square (2021)
'Federated Learning' (FL) is a method to train Artificial Intelligence (AI) models with data from multiple sources while maintaining anonymity of the data thus removing many barriers to data sharing. During the SARS-COV-2 pandemic, 20 institutes collaborated on a healthcare FL study to predict future oxygen requirements of infected patients using inputs of vital signs, laboratory data, and chest x-rays, constituting the "EXAM" (EMR CXR AI Model) model. EXAM achieved an average Area Under the Curve (AUC) of over 0.92, an average improvement of 16%, and a 38% increase in generalisability over local models. The FL paradigm was successfully applied to facilitate a rapid data science collaboration without data exchange, resulting in a model that generalised across heterogeneous, unharmonized datasets. This provided the broader healthcare community with a validated model to respond to COVID-19 challenges, as well as set the stage for broader use of FL in healthcare.