HQDSA-Seg

The HQDSA-Seg dataset contains 40 DSA sequences (160 images) from 110 Moyamoya disease patients, curated in collaboration with Nanjing Gulou Hospital. It features expert-annotated segmentation masks highlighting fine vascular structures and challenging \smoke-like\ vessels. A hybrid annotation pipeline was employed: images were first enhanced (\u03b1=1.5, \u03b2=0.5), filtered (Frangi filtering proved superior for complex vasculature), and binarized; results were then manually refined using Procreate software. This approach specifically addresses the segmentation difficulties posed by Moyamoya morphology.