Comparison of Privacy-Preserving Distributed Deep Learning Methods in Healthcare
This research addresses the challenge of privacy-preserving distributed deep learning in healthcare, specifically for medical image analysis, which is an incremental improvement to existing methods.
This paper compares federated learning, split learning, and SplitFed for detecting tuberculosis from chest X-rays. They propose a new architecture, SplitFedv3, which outperforms SplitFedv2 and split learning, and an alternate mini-batch training technique for split learning that improves performance over alternate client training.
In this paper, we compare three privacy-preserving distributed learning techniques: federated learning, split learning, and SplitFed. We use these techniques to develop binary classification models for detecting tuberculosis from chest X-rays and compare them in terms of classification performance, communication and computational costs, and training time. We propose a novel distributed learning architecture called SplitFedv3, which performs better than split learning and SplitFedv2 in our experiments. We also propose alternate mini-batch training, a new training technique for split learning, that performs better than alternate client training, where clients take turns to train a model.