<p>Effective and automated measurement of coronary lesions is essential for timely decision-making during interventions. However, a comprehensive, real-time strategy remains limited. This study aimed to develop a real-time deep learning system for automated detection and quantification of stenotic lesions in coronary angiography. The model was trained using 2651 diagnostic coronary angiographic images from 502 adult patients collected between February 2015 and January 2022 at two tertiary care hospitals. The system integrates five core components: vessel type classification, keyframe selection, lesion detection, vessel segmentation, and quantitative coronary angiography (QCA). In internal and external datasets, vessel type classification accuracies reached 96.33% and 94.19%, while keyframe selection accuracies were 98.29% and 93.27%, respectively. Lesion detection achieved recall/precision scores of 0.93/0.89 internally and 0.92/0.76 externally. Segmentation and QCA accuracies exceeded 0.92 in both cohorts. The complete system identifies stenotic lesions and their locations within 2 min. Clinical feedback indicated over 80% satisfaction. Our findings support the potential of this model to improve diagnostic accuracy and streamline clinical workflows in coronary angiography.</p> Graphical Abstract <p>The summary of steps of data process. The first model generates auto-DICOM classification. The second model is keyframe selections. The third model generates auto-lesion detections. The fourth model generates vessel segmentation. The fifth model includes automatic quantitative coronary angiography (QCA) and post processing. The final output report comprises information including vessel types, severity of stenosis, lesion width, and length and locations. LAD = left anterior descending artery, LCX = left circumflex artery, and RCA = right coronary artery</p> <p></p>