Fair Contribution Evaluation in Federated Learning for Medical Image Analysis Using the Banzhaf Index

This paper explores the application of Federated Learning (FL) in medical image analysis, with a focus on how the Banzhaf Index can be used to assess participants' contributions. Federated Learning, as a distributed machine learning approach, allows for collaborative training while preserving data privacy by keeping data localized to participants' devices. However, data heterogeneity and the imbalance in contribution evaluation remain major challenges in FL. This study proposes a Federated Learning model based on the Banzhaf Index, which quantifies each participant's contribution to the global model by calculating their marginal contributions. Experimental results demonstrate that the Banzhaf Index effectively detects malicious participants and provides a fair contribution evaluation in scenarios with imbalanced data distributions. This research presents a novel approach and method for applying Federated Learning in medical image analysis.