Federated Self-Supervised Learning for One-Shot Cross-Modal and Cross-Imaging Technique Segmentation
Siladittya Manna, Suresh Das, Sayantari Ghosh, Saumik Bhattacharya
TL;DR
The paper tackles privacy-preserving, cross-modal one-shot segmentation in a federated setting by adapting CoWPro into FedCoWPro, a self-supervised framework that learns representations across heterogeneous MR/CT data. It introduces two Dice-based regularizers—Spatial Dice Loss and Edge Dice Loss—and defines a final objective $L_{total} = L_{cowpro} + L_{ds} + L_{es}$ to boost segmentation accuracy without fine-tuning on downstream tasks. A novel public brachytherapy MR dataset (MOGaMB) is introduced alongside established public datasets (CHAOS, BTCV, FLARE22) to enable held-out evaluation and cross-domain testing. Results indicate that FedCoWPro can match or surpass FedAvg-based CoWPro baselines on unseen validation data, highlighting the feasibility and practical impact of federated self-supervised one-shot segmentation in privacy-sensitive medical imaging. The work thus paves the way for robust, low-annotation, privacy-preserving segmentation across institutions and imaging modalities.
Abstract
Decentralized federated learning enables learning of data representations from multiple sources without compromising the privacy of the clients. In applications like medical image segmentation, where obtaining a large annotated dataset from a single source is a distressing problem, federated self-supervised learning can provide some solace. In this work, we push the limits further by exploring a federated self-supervised one-shot segmentation task representing a more data-scarce scenario. We adopt a pre-existing self-supervised few-shot segmentation framework CoWPro and adapt it to the federated learning scenario. To the best of our knowledge, this work is the first to attempt a self-supervised few-shot segmentation task in the federated learning domain. Moreover, we consider the clients to be constituted of data from different modalities and imaging techniques like MR or CT, which makes the problem even harder. Additionally, we reinforce and improve the baseline CoWPro method using a fused dice loss which shows considerable improvement in performance over the baseline CoWPro. Finally, we evaluate this novel framework on a completely unseen held-out part of the local client dataset. We observe that the proposed framework can achieve performance at par or better than the FedAvg version of the CoWPro framework on the held-out validation dataset.