Federated Vision Transformer with Adaptive Focal Loss for Medical Image Classification

2026-02-01

Knowledge-Based Systems

AI Summary

This work addresses the challenge of limited model generalization in federated learning caused by data heterogeneity and class imbalance in medical imaging. To this end, we propose a novel federated learning framework that integrates Vision Transformers, Dynamic Adaptive Focal Loss (DAFL), and client-aware weighted aggregation. The DAFL dynamically adjusts loss weights based on an adaptive class imbalance coefficient, while the aggregation strategy is tailored to each client’s local data distribution, thereby enhancing attention to minority classes and improving cross-client generalization—all while preserving data privacy. Extensive experiments on the ISIC, Ocular Disease, and RSNA-ICH datasets demonstrate that our approach consistently outperforms state-of-the-art methods, achieving accuracy improvements ranging from 0.98% to 41.69%.

Technology Category

• Machine Learning: Distributed Machine Learning & Federated Learning
• Computer Vision: Large Vision Models
• Humans and AI: Human-in-the-loop Machine Learning

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: Web learning to rank, online learning, and counterfactual learning for ranking

Problem

• Federated Learning
• Medical Image Classification
• Class Imbalance
• Data Heterogeneity
• Vision Transformer

Innovation

• Federated Learning
• Vision Transformer
• Adaptive Focal Loss
• Class Imbalance
• Client Heterogeneity