MS Lesion Segmentation: Revisiting Weighting Mechanisms for Federated Learning
This work addresses the challenge of scanner and acquisition variance in medical image analysis for MS lesion segmentation, offering an incremental improvement in federated learning methods.
The paper tackled the problem of multiple sclerosis lesion segmentation in federated learning by introducing two re-weighting mechanisms for local node aggregation and loss function, resulting in significant outperformance over other FL methods and even exceeding centralized training performance.
Federated learning (FL) has been widely employed for medical image analysis to facilitate multi-client collaborative learning without sharing raw data. Despite great success, FL's performance is limited for multiple sclerosis (MS) lesion segmentation tasks, due to variance in lesion characteristics imparted by different scanners and acquisition parameters. In this work, we propose the first FL MS lesion segmentation framework via two effective re-weighting mechanisms. Specifically, a learnable weight is assigned to each local node during the aggregation process, based on its segmentation performance. In addition, the segmentation loss function in each client is also re-weighted according to the lesion volume for the data during training. Comparison experiments on two FL MS segmentation scenarios using public and clinical datasets have demonstrated the effectiveness of the proposed method by outperforming other FL methods significantly. Furthermore, the segmentation performance of FL incorporating our proposed aggregation mechanism can exceed centralised training with all the raw data. The extensive evaluation also indicated the superiority of our method when estimating brain volume differences estimation after lesion inpainting.