84 231

Cited 29 times in

Federated Learning for Thyroid Ultrasound Image Analysis to Protect Personal Information: Validation Study in a Real Health Care Environment

DC Field Value Language
dc.contributor.author김석모-
dc.date.accessioned2022-09-14T01:20:36Z-
dc.date.available2022-09-14T01:20:36Z-
dc.date.issued2021-05-
dc.identifier.urihttps://ir.ymlib.yonsei.ac.kr/handle/22282913/190408-
dc.description.abstractBackground: Federated learning is a decentralized approach to machine learning; it is a training strategy that overcomes medical data privacy regulations and generalizes deep learning algorithms. Federated learning mitigates many systemic privacy risks by sharing only the model and parameters for training, without the need to export existing medical data sets. In this study, we performed ultrasound image analysis using federated learning to predict whether thyroid nodules were benign or malignant. Objective: The goal of this study was to evaluate whether the performance of federated learning was comparable with that of conventional deep learning. Methods: A total of 8457 (5375 malignant, 3082 benign) ultrasound images were collected from 6 institutions and used for federated learning and conventional deep learning. Five deep learning networks (VGG19, ResNet50, ResNext50, SE-ResNet50, and SE-ResNext50) were used. Using stratified random sampling, we selected 20% (1075 malignant, 616 benign) of the total images for internal validation. For external validation, we used 100 ultrasound images (50 malignant, 50 benign) from another institution. Results: For internal validation, the area under the receiver operating characteristic (AUROC) curve for federated learning was between 78.88% and 87.56%, and the AUROC for conventional deep learning was between 82.61% and 91.57%. For external validation, the AUROC for federated learning was between 75.20% and 86.72%, and the AUROC curve for conventional deep learning was between 73.04% and 91.04%. Conclusions: We demonstrated that the performance of federated learning using decentralized data was comparable to that of conventional deep learning using pooled data. Federated learning might be potentially useful for analyzing medical images while protecting patients' personal information.-
dc.description.statementOfResponsibilityopen-
dc.languageEnglish-
dc.publisherJMIR Publications-
dc.relation.isPartOfJMIR MEDICAL INFORMATICS-
dc.rightsCC BY-NC-ND 2.0 KR-
dc.titleFederated Learning for Thyroid Ultrasound Image Analysis to Protect Personal Information: Validation Study in a Real Health Care Environment-
dc.typeArticle-
dc.contributor.collegeCollege of Medicine (의과대학)-
dc.contributor.departmentDept. of Surgery (외과학교실)-
dc.contributor.googleauthorHaeyun Lee-
dc.contributor.googleauthorYoung Jun Chai-
dc.contributor.googleauthorHyunjin Joo-
dc.contributor.googleauthorKyungsu Lee-
dc.contributor.googleauthorJae Youn Hwang-
dc.contributor.googleauthorSeok-Mo Kim-
dc.contributor.googleauthorKwangsoon Kim-
dc.contributor.googleauthorInn-Chul Nam-
dc.contributor.googleauthorJune Young Choi-
dc.contributor.googleauthorHyeong Won Yu-
dc.contributor.googleauthorMyung-Chul Lee-
dc.contributor.googleauthorHiroo Masuoka-
dc.contributor.googleauthorAkira Miyauchi-
dc.contributor.googleauthorKyu Eun Lee-
dc.contributor.googleauthorSungwan Kim-
dc.contributor.googleauthorHyoun-Joong Kong-
dc.identifier.doi10.2196/25869-
dc.contributor.localIdA00542-
dc.relation.journalcodeJ03664-
dc.identifier.eissn2291-9694-
dc.identifier.pmid33858817-
dc.subject.keyworddeep learning-
dc.subject.keywordfederated learning-
dc.subject.keywordthyroid nodules-
dc.subject.keywordultrasound image-
dc.contributor.alternativeNameKim, Seok Mo-
dc.contributor.affiliatedAuthor김석모-
dc.citation.volume9-
dc.citation.number5-
dc.citation.startPagee25869-
dc.identifier.bibliographicCitationJMIR MEDICAL INFORMATICS, Vol.9(5) : e25869, 2021-05-
Appears in Collections:
1. College of Medicine (의과대학) > Dept. of Surgery (외과학교실) > 1. Journal Papers

qrcode

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.