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

被引：41

作者：

Lee, Haeyun ^{[1
,2
]}

Chai, Young Jun ^{[3
]}

Joo, Hyunjin ^{[1
,4
]}

Lee, Kyungsu ^{[1
,2
]}

Hwang, Jae Youn ^{[2
]}

Kim, Seok-Mo ^{[5
]}

Kim, Kwangsoon ^{[6
]}

Nam, Inn-Chul ^{[7
]}

Choi, June Young ^{[8
]}

Yu, Hyeong Won ^{[8
]}

Lee, Myung-Chul ^{[9
]}

Masuoka, Hiroo ^{[10
]}

Miyauchi, Akira ^{[10
]}

Lee, Kyu Eun ^{[1
,11
,12
]}

Kim, Sungwan ^{[1
,4
,13
]}

Kong, Hyoun-Joong ^{[1
,4
,14
]}

机构：

[1] Seoul Natl Univ, Med Res Ctr, Inst Med & Biol Engn, Coll Med, Seoul, South Korea

[2] Daegu Gyeongbuk Inst Sci & Technol, Dept Informat & Commun Engn, Daegu, South Korea

[3] Seoul Metropolitan Govt Seoul Natl Univ, Dept Surg, Boramae Med Ctr, Seoul, South Korea

[4] Seoul Natl Univ Hosp, Transdisciplinary Dept Med & Adv Technol, Daehak Ro 101, Seoul, South Korea

[5] Gangnam Severance Hosp, Thyroid Canc Ctr, Dept Surg, Seoul, South Korea

[6] Catholic Univ Korea, Coll Med, Dept Surg, Seoul, South Korea

[7] Catholic Univ Korea, Coll Med, Dept Otolaryngol Head & Neck Surg, Seoul, South Korea

[8] Seoul Natl Univ, Dept Surg, Bundang Hosp, Seongnam Si, Gyeonggi Do, South Korea

[9] Korea Canc Ctr Hosp, Korea Inst Radiol & Med Sci, Dept Otorhinolaryngol Head & Neck Surg, Seoul, South Korea

[10] Kuma Hosp, Dept Surg, Kobe, Hyogo, Japan

[11] Seoul Natl Univ Hosp, Dept Surg, Seoul, South Korea

[12] Coll Med, Seoul, South Korea

[13] Seoul Natl Univ, Dept Biomed Engn, Coll Med, Seoul, South Korea

[14] Seoul Natl Univ, Dept Med, Coll Med, Seoul, South Korea

来源：

JMIR MEDICAL INFORMATICS | 2021年 / 9卷 / 05期

关键词：

deep learning; federated learning; thyroid nodules; ultrasound image; CANCER; CLASSIFICATION;

D O I：

10.2196/25869

中图分类号：

R-058 [];

学科分类号：

摘要：

Background: 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.

引用

页数：9

共 30 条

[1] Breast cancer classification using deep belief networks [J].