Med-DualLoRA: Local Adaptation of Foundation Models for 3D Cardiac MRI

2026-03-11

AI Summary

This work addresses the challenge of 3D cardiac MRI-based disease detection in a multi-center, non-IID, and privacy-constrained setting by proposing a parameter-efficient federated learning framework. It introduces, for the first time in medical federated learning, a dual-branch Low-Rank Adaptation (LoRA) mechanism that decomposes LoRA additively into global shared and local private components, aggregating only the global part to decouple global knowledge sharing from local feature preservation. By fine-tuning just two Transformer modules, the method substantially reduces communication overhead while enhancing personalization. Evaluated on the ACDC and M&Ms datasets, the model achieves a balanced accuracy of 0.768 and a specificity of 0.612, significantly outperforming existing federated parameter-efficient fine-tuning approaches while maintaining communication efficiency.

Technology Category

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

Application Category

• User Modeling, Personalization and Recommendation: Federated recommendation systems and personalization
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Systems and Infrastructure for Web, Mobile and WoT: Federated Web and WoT systems, including distributed, federated and edge-based data processing

Abstract

Foundation models (FMs) show great promise for robust downstream performance across medical imaging tasks and modalities, including cardiac magnetic resonance (CMR), following task-specific adaptation. However, adaptation using single-site data may lead to suboptimal performance and increased model bias, while centralized fine-tuning on clinical data is often infeasible due to privacy constraints. Federated fine-tuning offers a privacy-preserving alternative; yet conventional approaches struggle under heterogeneous, non-IID multi-center data and incur substantial communication overhead when adapting large models. In this work, we study federated FM fine-tuning for 3D CMR disease detection and propose Med-DualLoRA, a client-aware parameter-efficient fine-tuning (PEFT) federated framework that disentangles globally shared and local low-rank adaptations (LoRA) through additive decomposition. Global and local LoRA modules are trained locally, but only the global component is shared and aggregated across sites, keeping local adapters private. This design improves personalization while significantly reducing communication cost, and experiments show that adapting only two transformer blocks preserves performance while further improving efficiency. We evaluate our method on a multi-center state-of-the-art cine 3D CMR FM fine-tuned for disease detection using ACDC and combined M\&Ms datasets, treating each vendor as a federated client. Med-DualLoRA achieves statistically significant improved performance (balanced accuracy 0.768, specificity 0.612) compared to other federated PEFT baselines, while maintaining communication efficiency. Our approach provides a scalable solution for local federated adaptation of medical FMs under realistic clinical constraints.

Problem

• federated learning
• foundation models
• 3D cardiac MRI
• non-IID data
• privacy-preserving

Innovation

• Federated Learning
• Parameter-Efficient Fine-Tuning
• LoRA
• Medical Foundation Models
• 3D Cardiac MRI