Quality-Adaptive Split-Federated Learning for Segmenting Medical Images with Inaccurate Annotations
This addresses the challenge of annotation variability in decentralized medical imaging, offering a practical solution for federated learning scenarios, though it is incremental as it builds on existing split-federated learning methods.
The paper tackles the problem of inaccurate and inconsistent annotations in medical image segmentation under split-federated learning by proposing QA-SplitFed, a data quality-based adaptive averaging strategy. Experiments on human embryo image segmentation show that QA-SplitFed maintains accuracy effectively with at least one uncorrupted client, while five baseline methods fail as corrupted clients increase.
SplitFed Learning, a combination of Federated and Split Learning (FL and SL), is one of the most recent developments in the decentralized machine learning domain. In SplitFed learning, a model is trained by clients and a server collaboratively. For image segmentation, labels are created at each client independently and, therefore, are subject to clients' bias, inaccuracies, and inconsistencies. In this paper, we propose a data quality-based adaptive averaging strategy for SplitFed learning, called QA-SplitFed, to cope with the variation of annotated ground truth (GT) quality over multiple clients. The proposed method is compared against five state-of-the-art model averaging methods on the task of learning human embryo image segmentation. Our experiments show that all five baseline methods fail to maintain accuracy as the number of corrupted clients increases. QA-SplitFed, however, copes effectively with corruption as long as there is at least one uncorrupted client.