Coronary artery segmentation is critical for diagnosing and planning treatment for cardiovascular diseases, yet it remains challenging in X-ray angiography due to the complexity of vessel structures and the difficulty in identifying small, low-contrast branches. To tackle these issues, this paper proposes a two-stage approach. The first stage, Vessel Segment Decomposition, utilizes a tracking-based strategy to decompose the vessel network into individual segments, effectively isolating small and intricate branches. The second stage, Semantic Enhancement Segmentation, refines segmentation by leveraging these segments to ensure semantic consistency and preserve vessel connectivity. Experiments on a dedicated clinical dataset show a 2.48% improvement in Dice score and a 22.54% increase in segment recall rate at a threshold of 0.8, validating the method’s robustness and generalizability.