Federated Learning for Breast Density Classification: A Real-World Implementation
Abstract
Building robust deep learning-based models requires large quantities of diverse training data. In this study, we investigate the use of federated learning (FL) to build medical imaging classification models in a real-world collaborative setting. Seven clinical institutions from across the world joined this FL effort to train a model for breast density classification based on Breast Imaging, Reporting & Data System (BI-RADS). We show that despite substantial differences among the datasets from all sites (mammography system, class distribution, and data set size) and without centralizing data, we can successfully train AI models in federation. The results show that models trained using FL perform 6.3% on average better than their counterparts trained on an institute's local data alone. Furthermore, we show a 45.8% relative improvement in the models' generalizability when evaluated on the other participating sites' testing data.
Keywords
Cite
@article{arxiv.2009.01871,
title = {Federated Learning for Breast Density Classification: A Real-World Implementation},
author = {Holger R. Roth and Ken Chang and Praveer Singh and Nir Neumark and Wenqi Li and Vikash Gupta and Sharut Gupta and Liangqiong Qu and Alvin Ihsani and Bernardo C. Bizzo and Yuhong Wen and Varun Buch and Meesam Shah and Felipe Kitamura and Matheus Mendonça and Vitor Lavor and Ahmed Harouni and Colin Compas and Jesse Tetreault and Prerna Dogra and Yan Cheng and Selnur Erdal and Richard White and Behrooz Hashemian and Thomas Schultz and Miao Zhang and Adam McCarthy and B. Min Yun and Elshaimaa Sharaf and Katharina V. Hoebel and Jay B. Patel and Bryan Chen and Sean Ko and Evan Leibovitz and Etta D. Pisano and Laura Coombs and Daguang Xu and Keith J. Dreyer and Ittai Dayan and Ram C. Naidu and Mona Flores and Daniel Rubin and Jayashree Kalpathy-Cramer},
journal= {arXiv preprint arXiv:2009.01871},
year = {2020}
}
Comments
Accepted at the 1st MICCAI Workshop on "Distributed And Collaborative Learning"; add citation to Fig. 1 & 2 and update Fig. 5; fix typo in affiliations