Accurate coronary artery vessel segmentation is an important step for diagnosis and treatment of coronary heart disease. However, X-ray coronary angiography (XCA) images have complex structure, blurred details, scarce information, and strong noise, thus existing methods are still faced with challenges in vascular structure continuity modeling, spatial relationship characterization, and edge information extraction. To address the problems above, this paper proposes a modified TransUNet network structure, called SATENet, which consists of two new modules, that is, the Structure-Aware Topological Module (SATM) and the Edge Enhancement Module (EEM). In particular, SATM explicitly models long-range dependencies and topological structure of vessels to help the model learn spatial continuity and vascular connectivity, which is improved in segmenting long and tortuous vessel regions. Consequently, EEM adopt the multi-scale edges to capture up the multi-scale edges to improve boundary localization and reduce the effect of low contrast and blurred edges. We evaluated SATENet on the ARCADE dataset, and the experimental results show that the model has advantages over traditional models in overall segmentation accuracy and vascular boundary detail retention, indicating strong potential for clinical application.