MFM-DA: Instance-Aware Adaptor and Hierarchical Alignment for Efficient Domain Adaptation in Medical Foundation Models
This work addresses the domain adaptation challenge for medical imaging practitioners, offering a practical solution to improve MFM deployment in real-world scenarios, though it is incremental as it builds on existing UDA and adaptation techniques.
The paper tackles the domain gap problem in Medical Foundation Models (MFMs) by proposing MFM-DA, a few-shot unsupervised domain adaptation framework that uses a dynamic instance-aware adaptor and hierarchical alignment to translate source-domain images to the target domain, achieving superior performance on optic cup and disc segmentation tasks compared to state-of-the-art methods.
Medical Foundation Models (MFMs), trained on large-scale datasets, have demonstrated superior performance across various tasks. However, these models still struggle with domain gaps in practical applications. Specifically, even after fine-tuning on source-domain data, task-adapted foundation models often perform poorly in the target domain. To address this challenge, we propose a few-shot unsupervised domain adaptation (UDA) framework for MFMs, named MFM-DA, which only leverages a limited number of unlabeled target-domain images. Our approach begins by training a Denoising Diffusion Probabilistic Model (DDPM), which is then adapted to the target domain using a proposed dynamic instance-aware adaptor and a distribution direction loss, enabling the DDPM to translate source-domain images into the target domain style. The adapted images are subsequently processed through the MFM, where we introduce a designed channel-spatial alignment Low-Rank Adaptation (LoRA) to ensure effective feature alignment. Extensive experiments on optic cup and disc segmentation tasks demonstrate that MFM-DA outperforms state-of-the-art methods. Our work provides a practical solution to the domain gap issue in real-world MFM deployment. Code will be available at here.