Addressing Annotation Scarcity in Hyperspectral Brain Image Segmentation with Unsupervised Domain Adaptation

This work presents a novel deep learning framework for segmenting cerebral vasculature in hyperspectral brain images. We address the critical challenge of severe label scarcity, which impedes conventional supervised training. Our approach utilizes a novel unsupervised domain adaptation methodology, completely avoiding the need for target-domain labels. The method segments the brain vasculature of the target domain using only labelled retinal images from a different dataset. Quantitative and qualitative evaluations confirm that our method significantly outperforms existing state-of-the-art approaches, demonstrating the efficacy of domain adaptation for label-scarce biomedical imaging tasks.