Robust Source-Free Domain Adaptation for Medical Image Segmentation based on Curriculum Learning
This addresses data privacy and security concerns in medical imaging by enabling domain adaptation without sharing sensitive source data, though it is incremental as it builds on existing source-free adaptation methods.
The paper tackles the problem of adapting medical image segmentation models to new domains without access to the original source data, using a curriculum learning approach that combines easy-to-hard and source-to-target curricula, achieving state-of-the-art results on fundus and polyp segmentation datasets.
Recent studies have uncovered a new research line, namely source-free domain adaptation, which adapts a model to target domains without using the source data. Such a setting can address the concerns on data privacy and security issues of medical images. However, current source-free domain adaptation frameworks mainly focus on the pseudo label refinement for target data without the consideration of learning procedure. Indeed, a progressive learning process from source to target domain will benefit the knowledge transfer during model adaptation. To this end, we propose a curriculum-based framework, namely learning from curriculum (LFC), for source-free domain adaptation, which consists of easy-to-hard and source-to-target curricula. Concretely, the former curriculum enables the framework to start learning with `easy' samples and gradually tune the optimization direction of model adaption by increasing the sample difficulty. While, the latter can stablize the adaptation process, which ensures smooth transfer of the model from the source domain to the target. We evaluate the proposed source-free domain adaptation approach on the public cross-domain datasets for fundus segmentation and polyp segmentation. The extensive experimental results show that our framework surpasses the existing approaches and achieves a new state-of-the-art.