Automated Pancreas Segmentation Using Multi-institutional Collaborative Deep Learning

TL;DR

This paper explores federated learning for medical image segmentation, demonstrating that collaborative training across institutions improves model generalizability without sharing sensitive data.

Contribution

It introduces a federated learning approach for pancreas segmentation, addressing data privacy issues and comparing its effectiveness to local training.

Findings

Federated learning improves segmentation model generalizability.

Collaborative training outperforms local-only models.

Effective in real-world multi-institutional settings.

Abstract

The performance of deep learning-based methods strongly relies on the number of datasets used for training. Many efforts have been made to increase the data in the medical image analysis field. However, unlike photography images, it is hard to generate centralized databases to collect medical images because of numerous technical, legal, and privacy issues. In this work, we study the use of federated learning between two institutions in a real-world setting to collaboratively train a model without sharing the raw data across national boundaries. We quantitatively compare the segmentation models obtained with federated learning and local training alone. Our experimental results show that federated learning models have higher generalizability than standalone training.