Divide, Conquer and Unite: Hierarchical Style-Recalibrated Prototype Alignment for Federated Medical Image Segmentation

2025-11-14

AI Summary

In federated medical image segmentation, inter-site variations in imaging equipment and acquisition protocols induce feature heterogeneity, degrading model generalizability. To address this, we propose a frequency-domain adaptive style recalibration and dual-level context-aware prototype alignment framework. Our method innovatively decouples content and style representations in the frequency domain to suppress cumulative style bias in intermediate network layers. Furthermore, we design an encoder–decoder cascaded domain-invariant context prototype alignment mechanism that integrates multi-scale semantic cues to enhance cross-site feature consistency. Extensive experiments on two public benchmarks demonstrate that our approach significantly improves segmentation accuracy and outperforms state-of-the-art federated segmentation methods in both convergence stability and generalization across heterogeneous clients.

Technology Category

• Computer Vision: Segmentation
• Machine Learning: Distributed Machine Learning & Federated Learning
• Constraint Satisfaction and Optimization: Distributed CSP/Optimization

Application Category

• User Modeling, Personalization and Recommendation: Federated recommendation systems and personalization
• Systems and Infrastructure for Web, Mobile and WoT: Federated Web and WoT systems, including distributed, federated and edge-based data processing
• Search and Retrieval-Augmented AI: Vertical and domain-specific search

Abstract

Federated learning enables multiple medical institutions to train a global model without sharing data, yet feature heterogeneity from diverse scanners or protocols remains a major challenge. Many existing works attempt to address this issue by leveraging model representations (e.g., mean feature vectors) to correct local training; however, they often face two key limitations: 1) Incomplete Contextual Representation Learning: Current approaches primarily focus on final-layer features, overlooking critical multi-level cues and thus diluting essential context for accurate segmentation. 2) Layerwise Style Bias Accumulation: Although utilizing representations can partially align global features, these methods neglect domain-specific biases within intermediate layers, allowing style discrepancies to build up and reduce model robustness. To address these challenges, we propose FedBCS to bridge feature representation gaps via domain-invariant contextual prototypes alignment. Specifically, we introduce a frequency-domain adaptive style recalibration into prototype construction that not only decouples content-style representations but also learns optimal style parameters, enabling more robust domain-invariant prototypes. Furthermore, we design a context-aware dual-level prototype alignment method that extracts domain-invariant prototypes from different layers of both encoder and decoder and fuses them with contextual information for finer-grained representation alignment. Extensive experiments on two public datasets demonstrate that our method exhibits remarkable performance.

Problem

• Addresses feature heterogeneity in federated medical image segmentation across institutions
• Overcomes incomplete contextual representation learning from single-layer features
• Mitigates layerwise style bias accumulation in intermediate network layers

Innovation

• Hierarchical style-recalibrated prototype alignment for segmentation
• Frequency-domain adaptive style recalibration in prototype construction
• Context-aware dual-level prototype alignment across network layers