Federated Continual 3D Segmentation With Single-round Communication

2025-03-19

AI Summary

In dynamic federated medical settings, continual client enrollment and incremental label-set expansion incur excessive communication overhead, frequent synchronization, and redundant retraining. To address these challenges, this paper proposes the first single-round-communication federated continual learning framework for 3D medical image segmentation. Methodologically, it introduces a one-time server aggregation mechanism that integrates asynchronous client collaboration with multi-model knowledge distillation, enabling continuous global model evolution during local training—thereby eliminating full-network retraining and periodic synchronization. Evaluated on multi-class abdominal CT segmentation using a 3D CNN architecture, the framework reduces communication rounds by ~70%, significantly lowers computational cost, supports seamless client scalability, and maintains state-of-the-art segmentation accuracy. The core contribution lies in the deep integration of federated learning, continual learning, and knowledge distillation—achieving, for the first time, dynamic label-set expansion and sustained model optimization within a single communication round.

Technology Category

• Machine Learning: Distributed Machine Learning & Federated Learning
• Computer Vision: Segmentation
• Multiagent Systems: Multiagent Learning

Application Category

• User Modeling, Personalization and Recommendation: Federated recommendation systems and personalization
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Federated learning seeks to foster collaboration among distributed clients while preserving the privacy of their local data. Traditionally, federated learning methods assume a fixed setting in which client data and learning objectives remain constant. However, in real-world scenarios, new clients may join, and existing clients may expand the segmentation label set as task requirements evolve. In such a dynamic federated analysis setup, the conventional federated communication strategy of model aggregation per communication round is suboptimal. As new clients join, this strategy requires retraining, linearly increasing communication and computation overhead. It also imposes requirements for synchronized communication, which is difficult to achieve among distributed clients. In this paper, we propose a federated continual learning strategy that employs a one-time model aggregation at the server through multi-model distillation. This approach builds and updates the global model while eliminating the need for frequent server communication. When integrating new data streams or onboarding new clients, this approach efficiently reuses previous client models, avoiding the need to retrain the global model across the entire federation. By minimizing communication load and bypassing the need to put unchanged clients online, our approach relaxes synchronization requirements among clients, providing an efficient and scalable federated analysis framework suited for real-world applications. Using multi-class 3D abdominal CT segmentation as an application task, we demonstrate the effectiveness of the proposed approach.

Problem

• Dynamic federated learning with evolving client data and objectives.
• Minimizing communication and computation overhead in federated systems.
• Efficient integration of new clients without retraining global models.

Innovation

• Single-round communication via multi-model distillation
• Efficient reuse of previous client models
• Minimized communication and synchronization requirements