FHBench: Towards Efficient and Personalized Federated Learning for Multimodal Healthcare

2025-04-15

AI Summary

To address the dual challenges of constrained computational resources and insufficient personalization in multimodal medical federated learning, this paper introduces FHBench—the first multimodal federated benchmark tailored to real-world clinical scenarios—covering four major diagnostic tasks: neurological, cardio-cerebrovascular, respiratory, and pathological analysis. We further propose EPFL, an Efficient Personalized Federated Learning framework, which innovatively integrates adaptive LoRA-based fine-tuning with a lightweight heterogeneous model aggregation mechanism. FHBench fills a critical gap by establishing the first standardized multimodal federated benchmark for healthcare applications. Empirical evaluation demonstrates that EPFL achieves an average accuracy improvement of 3.2% across tasks, reduces communication overhead by 67%, and decreases inference latency by 58%, thereby simultaneously enhancing model personalization, communication efficiency, and deployment feasibility in resource-constrained clinical settings.

Technology Category

• Machine Learning: Multimodal Learning
• Natural Language Processing: Language Grounding & Multi-modal NLP
• Computer Vision: Multi-modal Vision

Application Category

• User Modeling, Personalization and Recommendation: Federated recommendation systems and personalization
• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

Federated Learning (FL) has emerged as an effective solution for multi-institutional collaborations without sharing patient data, offering a range of methods tailored for diverse applications. However, real-world medical datasets are often multimodal, and computational resources are limited, posing significant challenges for existing FL approaches. Recognizing these limitations, we developed the Federated Healthcare Benchmark(FHBench), a benchmark specifically designed from datasets derived from real-world healthcare applications. FHBench encompasses critical diagnostic tasks across domains such as the nervous, cardiovascular, and respiratory systems and general pathology, providing comprehensive support for multimodal healthcare evaluations and filling a significant gap in existing benchmarks. Building on FHBench, we introduced Efficient Personalized Federated Learning with Adaptive LoRA(EPFL), a personalized FL framework that demonstrates superior efficiency and effectiveness across various healthcare modalities. Our results highlight the robustness of FHBench as a benchmarking tool and the potential of EPFL as an innovative approach to advancing healthcare-focused FL, addressing key limitations of existing methods.

Problem

• Addressing multimodal healthcare data challenges in Federated Learning
• Developing a benchmark for real-world healthcare FL applications
• Proposing efficient personalized FL for diverse medical modalities

Innovation

• Federated Healthcare Benchmark for multimodal evaluations
• Efficient Personalized FL with Adaptive LoRA
• Supports nervous, cardiovascular, respiratory diagnostic tasks