{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,5,7]],"date-time":"2026-05-07T21:19:34Z","timestamp":1778188774007,"version":"3.51.4"},"reference-count":26,"publisher":"Oxford University Press (OUP)","issue":"5","license":[{"start":{"date-parts":[[2020,3,20]],"date-time":"2020-03-20T00:00:00Z","timestamp":1584662400000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/academic.oup.com\/journals\/pages\/open_access\/funder_policies\/chorus\/standard_publication_model"}],"funder":[{"DOI":"10.13039\/100000054","name":"National Cancer Institute","doi-asserted-by":"publisher","award":["U01CA242879"],"award-info":[{"award-number":["U01CA242879"]}],"id":[{"id":"10.13039\/100000054","id-type":"DOI","asserted-by":"publisher"}]},{"DOI":"10.13039\/100000070","name":"National Institute of Biomedical Imaging and Bioengineering","doi-asserted-by":"publisher","award":["5T32EB1680"],"award-info":[{"award-number":["5T32EB1680"]}],"id":[{"id":"10.13039\/100000070","id-type":"DOI","asserted-by":"publisher"}]},{"DOI":"10.13039\/100000054","name":"National Cancer Institute","doi-asserted-by":"publisher","award":["F30CA239407"],"award-info":[{"award-number":["F30CA239407"]}],"id":[{"id":"10.13039\/100000054","id-type":"DOI","asserted-by":"publisher"}]},{"DOI":"10.13039\/100000002","name":"National Institutes of Health","doi-asserted-by":"publisher","award":["U01CA154601"],"award-info":[{"award-number":["U01CA154601"]}],"id":[{"id":"10.13039\/100000002","id-type":"DOI","asserted-by":"publisher"}]},{"DOI":"10.13039\/100000002","name":"National Institutes of Health","doi-asserted-by":"publisher","award":["U24CA180927"],"award-info":[{"award-number":["U24CA180927"]}],"id":[{"id":"10.13039\/100000002","id-type":"DOI","asserted-by":"publisher"}]},{"DOI":"10.13039\/100000002","name":"National Institutes of Health","doi-asserted-by":"publisher","award":["U24CA180918"],"award-info":[{"award-number":["U24CA180918"]}],"id":[{"id":"10.13039\/100000002","id-type":"DOI","asserted-by":"publisher"}]},{"name":"Stanford AstraZeneca Collaboration program"},{"DOI":"10.13039\/100000070","name":"National Institute of Biomedical Imaging and Bioengineering","doi-asserted-by":"publisher","award":["P41EB015896"],"award-info":[{"award-number":["P41EB015896"]}],"id":[{"id":"10.13039\/100000070","id-type":"DOI","asserted-by":"publisher"}]},{"DOI":"10.13039\/100000002","name":"National Institutes of Health","doi-asserted-by":"publisher","id":[{"id":"10.13039\/100000002","id-type":"DOI","asserted-by":"publisher"}]}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":[],"published-print":{"date-parts":[[2020,5,1]]},"abstract":"<jats:title>Abstract<\/jats:title>\n               <jats:sec>\n                  <jats:title>Objectives<\/jats:title>\n                  <jats:p>Sharing patient data across institutions to train generalizable deep learning models is challenging due to regulatory and technical hurdles. Distributed learning, where model weights are shared instead of patient data, presents an attractive alternative. Cyclical weight transfer (CWT) has recently been demonstrated as an effective distributed learning method for medical imaging with homogeneous data across institutions. In this study, we optimize CWT to overcome performance losses from variability in training sample sizes and label distributions across institutions.<\/jats:p>\n               <\/jats:sec>\n               <jats:sec>\n                  <jats:title>Materials and Methods<\/jats:title>\n                  <jats:p>Optimizations included proportional local training iterations, cyclical learning rate, locally weighted minibatch sampling, and cyclically weighted loss. We evaluated our optimizations on simulated distributed diabetic retinopathy detection and chest radiograph classification.<\/jats:p>\n               <\/jats:sec>\n               <jats:sec>\n                  <jats:title>Results<\/jats:title>\n                  <jats:p>Proportional local training iteration mitigated performance losses from sample size variability, achieving 98.6% of the accuracy attained by centrally hosting in the diabetic retinopathy dataset split with highest sample size variance across institutions. Locally weighted minibatch sampling and cyclically weighted loss both mitigated performance losses from label distribution variability, achieving 98.6% and 99.1%, respectively, of the accuracy attained by centrally hosting in the diabetic retinopathy dataset split with highest label distribution variability across institutions.<\/jats:p>\n               <\/jats:sec>\n               <jats:sec>\n                  <jats:title>Discussion<\/jats:title>\n                  <jats:p>Our optimizations to CWT improve its capability of handling data variability across institutions. Compared to CWT without optimizations, CWT with optimizations achieved performance significantly closer to performance from centrally hosting.<\/jats:p>\n               <\/jats:sec>\n               <jats:sec>\n                  <jats:title>Conclusion<\/jats:title>\n                  <jats:p>Our work is the first to identify and address challenges of sample size and label distribution variability in simulated distributed deep learning for medical imaging. Future work is needed to address other sources of real-world data variability.<\/jats:p>\n               <\/jats:sec>","DOI":"10.1093\/jamia\/ocaa017","type":"journal-article","created":{"date-parts":[[2020,2,7]],"date-time":"2020-02-07T20:16:33Z","timestamp":1581106593000},"page":"700-708","source":"Crossref","is-referenced-by-count":52,"title":["Accounting for data variability in multi-institutional distributed deep learning for medical imaging"],"prefix":"10.1093","volume":"27","author":[{"ORCID":"https:\/\/orcid.org\/0000-0002-8961-2542","authenticated-orcid":false,"given":"Niranjan","family":"Balachandar","sequence":"first","affiliation":[{"name":"Laboratory of Quantitative Imaging and Artificial Intelligence, Department of Radiology and Biomedical Data Science, Stanford University, Stanford, CA, USA"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Ken","family":"Chang","sequence":"additional","affiliation":[{"name":"Athinoula A. Martinos Center for Biomedical Imaging, Department of Radiology, Massachusetts General Hospital, Charlestown, MA, USA"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Jayashree","family":"Kalpathy-Cramer","sequence":"additional","affiliation":[{"name":"Athinoula A. Martinos Center for Biomedical Imaging, Department of Radiology, Massachusetts General Hospital, Charlestown, MA, USA"},{"name":"MGH and BWH Center for Clinical Data Science, Massachusetts General Hospital, Boston, MA, USA"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Daniel L","family":"Rubin","sequence":"additional","affiliation":[{"name":"Laboratory of Quantitative Imaging and Artificial Intelligence, Department of Radiology and Biomedical Data Science, Stanford University, Stanford, CA, USA"}],"role":[{"role":"author","vocabulary":"crossref"}]}],"member":"286","published-online":{"date-parts":[[2020,3,20]]},"reference":[{"key":"2020110613100997900_ocaa017-B1","author":"Szegedy","year":"4\u20139, 2017;  ."},{"key":"2020110613100997900_ocaa017-B2","first-page":"91","article-title":"Faster r-cnn: towards real-time object detection with region proposal networks","author":"Ren","year":"2015","journal-title":"Adv Neural Inf Process Syst"},{"issue":"22","key":"2020110613100997900_ocaa017-B3","doi-asserted-by":"crossref","first-page":"2402","DOI":"10.1001\/jama.2016.17216","article-title":"Development and validation of a deep learning algorithm for detection of diabetic retinopathy in retinal fundus photographs","volume":"316","author":"Gulshan","year":"2016","journal-title":"JAMA"},{"issue":"7","key":"2020110613100997900_ocaa017-B4","doi-asserted-by":"crossref","first-page":"803","DOI":"10.1001\/jamaophthalmol.2018.1934","article-title":"Automated diagnosis of plus disease in retinopathy of prematurity using deep convolutional neural networks","volume":"136","author":"Brown","year":"2018","journal-title":"JAMA Ophthalmol"},{"issue":"5","key":"2020110613100997900_ocaa017-B5","doi-asserted-by":"crossref","first-page":"1160","DOI":"10.1109\/TMI.2016.2536809","article-title":"Pulmonary nodule detection in CT images: false positive reduction using multi-view convolutional networks","volume":"35","author":"Setio","year":"2016","journal-title":"IEEE Trans Med Imaging"},{"key":"2020110613100997900_ocaa017-B6","author":"Bakas","year":"2018"},{"issue":"11","key":"2020110613100997900_ocaa017-B7","doi-asserted-by":"crossref","first-page":"1412","DOI":"10.1093\/neuonc\/noz106","article-title":"Automatic assessment of glioma burden: A deep learning algorithm for fully automated volumetric and bi-dimensional measurement","volume":"21","author":"Chang","year":"2019","journal-title":"Neuro-Oncology"},{"issue":"11","key":"2020110613100997900_ocaa017-B8","doi-asserted-by":"crossref","first-page":"e1002683","DOI":"10.1371\/journal.pmed.1002683","article-title":"Variable generalization performance of a deep learning model to detect pneumonia in chest radiographs: a cross-sectional study","volume":"15","author":"Zech","year":"2018","journal-title":"PLOS Med"},{"issue":"3","key":"2020110613100997900_ocaa017-B9","doi-asserted-by":"crossref","first-page":"1150","DOI":"10.1002\/mp.12752","article-title":"Deep learning for segmentation of brain tumors: impact of cross-institutional training and testing","volume":"45","author":"AlBadawy","year":"2018","journal-title":"Med Phys"},{"issue":"8","key":"2020110613100997900_ocaa017-B10","doi-asserted-by":"crossref","first-page":"945","DOI":"10.1093\/jamia\/ocy017","article-title":"Distributed deep learning networks among institutions for medical imaging","volume":"25","author":"Chang","year":"2018","journal-title":"J Am Med Inform Assoc"},{"key":"2020110613100997900_ocaa017-B11","author":"Kone\u010dny`","year":"2016"},{"key":"2020110613100997900_ocaa017-B12","author":"Chen","year":"2016"},{"issue":"3","key":"2020110613100997900_ocaa017-B13","doi-asserted-by":"crossref","first-page":"781","DOI":"10.1148\/radiol.2019190613","article-title":"A roadmap for foundational research on artificial intelligence in medical imaging: from the 2018 NIH\/RSNA\/ACR\/The Academy Workshop","volume":"291","author":"Langlotz","year":"2019","journal-title":"Radiology"},{"key":"2020110613100997900_ocaa017-B14","first-page":"464","author":"Smith","year":"2017"},{"issue":"10","key":"2020110613100997900_ocaa017-B15","doi-asserted-by":"crossref","first-page":"3738","DOI":"10.1016\/j.patcog.2012.03.014","article-title":"Inverse random under sampling for class imbalance problem and its application to multi-label classification","volume":"45","author":"Tahir","year":"2012","journal-title":"Pattern Recognit"},{"issue":"2","key":"2020110613100997900_ocaa017-B16","doi-asserted-by":"crossref","first-page":"224","DOI":"10.7763\/IJMLC.2013.V3.307","article-title":"Addressing the class imbalance problem in medical datasets","volume":"3","author":"Rahman","year":"2013","journal-title":"Int J Mach Learn Comput"},{"key":"2020110613100997900_ocaa017-B17","first-page":"240","author":"Sudre","year":"2017"},{"key":"2020110613100997900_ocaa017-B18","year":"2015"},{"key":"2020110613100997900_ocaa017-B19","first-page":"2097","author":"Wang","year":"2017"},{"key":"2020110613100997900_ocaa017-B20","author":"Graham","year":"2015"},{"key":"2020110613100997900_ocaa017-B21","first-page":"1","author":"Szegedy","year":"2015"},{"key":"2020110613100997900_ocaa017-B22","author":"Kingma","year":"2014"},{"key":"2020110613100997900_ocaa017-B23","first-page":"770","author":"He","year":"2016"},{"key":"2020110613100997900_ocaa017-B24","first-page":"4681","author":"Ledig","year":"2017"},{"key":"2020110613100997900_ocaa017-B25","author":"Su","year":"2015"},{"key":"2020110613100997900_ocaa017-B26","author":"Vepakomma","year":"2018"}],"container-title":["Journal of the American Medical Informatics Association"],"original-title":[],"language":"en","link":[{"URL":"http:\/\/academic.oup.com\/jamia\/article-pdf\/27\/5\/700\/34152763\/ocaa017.pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"syndication"},{"URL":"http:\/\/academic.oup.com\/jamia\/article-pdf\/27\/5\/700\/34152763\/ocaa017.pdf","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2020,11,6]],"date-time":"2020-11-06T19:32:01Z","timestamp":1604691121000},"score":1,"resource":{"primary":{"URL":"https:\/\/academic.oup.com\/jamia\/article\/27\/5\/700\/5810515"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2020,3,20]]},"references-count":26,"journal-issue":{"issue":"5","published-online":{"date-parts":[[2020,3,20]]},"published-print":{"date-parts":[[2020,5,1]]}},"URL":"https:\/\/doi.org\/10.1093\/jamia\/ocaa017","relation":{},"ISSN":["1527-974X"],"issn-type":[{"value":"1527-974X","type":"electronic"}],"subject":[],"published-other":{"date-parts":[[2020,5]]},"published":{"date-parts":[[2020,3,20]]}}}