Coronary Artery Segmentation from X-ray Angiographic Images using Width-aware U-Net

Shuang Song,Ruoxiu Xiao,Tong Lu,Hang Zhu

DOI: https://doi.org/10.1109/ICCGIV57403.2022.00028

2022-01-01

Abstract:X-ray angiographic images are regarded as the gold standard in the intra-operation of vascular diseases due to their high imaging quality and speed. However, due to many vascular- similar structures existing in the images, the accurate segmentation of the coronary artery is challenging. In this paper, we proposed a width-aware U-Net model by supervising the segmentation procedure using the label mask and centerline- derived width map. In our model, the centerlines of label masks are extracted and width maps are then computed. Together with the label mask, the width maps are then utilized to supervise the segmentation network. By comparing the proposed network with other state-of-the-art models, we can achieve superior results in discarding non-vascular artefacts and preserving the integrity of the whole vascular tree.