Institutionally Distributed Deep Learning Networks
This addresses the challenge of collaborative medical diagnosis for institutions with data privacy concerns, but it is incremental as it builds on existing distributed learning methods.
The study tackled the problem of training deep learning models across multiple institutions without sharing patient data by simulating model dissemination with various heuristics, finding that cyclical weight transfer achieved a testing accuracy of 77.3%, close to the 78.7% accuracy of centrally trained models.
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.