Institutionally Distributed Deep Learning Networks

TL;DR

This paper compares different distributed deep learning methods for medical image classification across institutions, finding cyclical weight transfer nearly matches centralized training performance.

Contribution

It introduces and evaluates heuristics for distributed deep learning across institutions, highlighting cyclical weight transfer as an effective approach.

Findings

Cyclical weight transfer achieves 77.3% accuracy, close to 78.7% of centralized data.

High frequency of weight transfer improves performance.

Ensembling and single weight transfer are less effective.

Abstract

Deep learning has become a promising approach for automated medical diagnoses. When medical data samples are limited, collaboration among multiple institutions is necessary to achieve high algorithm performance. However, sharing patient data often has limitations due to technical, legal, or ethical concerns. In such cases, sharing a deep learning model is a more attractive alternative. The best method of performing such a task is unclear, however. In this study, we simulate the dissemination of learning deep learning network models across four institutions using various heuristics and compare the results with a deep learning model trained on centrally hosted patient data. The heuristics investigated include ensembling single institution models, single weight transfer, and cyclical weight transfer. We evaluated these approaches for image classification in three independent image collections (retinal fundus photos, mammography, and ImageNet). We find that cyclical weight transfer resulted in a performance (testing accuracy = 77.3%) that was closest to that of centrally hosted patient data (testing accuracy = 78.7%). We also found that there is an improvement in the performance of cyclical weight transfer heuristic with high frequency of weight transfer.