The automatic segmencrucial for cardiovascular diagnostics, aiding in the detection and monitoring of vascular abnormalities. In the present chapter, a novel method based on the U-Net architecture is proposed for the segmentation and skeletonization of coronary arteries. The method is structured in two main stages: segmentation and skeletonization, utilizing a U-Net architecture with a VGG16 backbone for feature extraction, and thresholding via inter-class variance method for image binarization. The proposed method was tested on a dataset of 134 X-ray coronary angiograms and compared with both classical segmentation and skeletonization methods, as well as various state-of-the-art skeletonization techniques. The performance was evaluated using accuracy and F1 score metrics. The segmentation model achieved an accuracy of 97.27% and an F1 score of 0.7202, outperforming the comparative techniques. The skeletonization model achieved an accuracy of 99.61% and an F1 score of 0.7797, surpassing classical methods. The proposed method shows promise for enhancing coronary artery analysis, offering a reliable tool for vascular health assessment, and potentially benefiting clinical applications requiring precise artery diameter measurements and detailed vessel structures.