This chapter provides a comprehensive survey of state-of-the-art deep learning (DL) architectures applied to the detection of coronary stenosis in X-ray angiograms. Given the critical role of early diagnosis in managing cardiovascular diseases, the chapter explores how various convolutional neural networks contribute to improving detection accuracy and efficiency. The focus is on key models such as SSD, R-CNN, RetinaNet, and different versions of YOLO, comparing their structures, methodologies, and suitability for real-time applications. The study is based on a labeled dataset composed of 902 images, enhanced by seven data augmentation operations to increase the diversity of positive case images from 69 to 552 (483 new images \(+\) 69 original images). Each architecture is evaluated using performance metrics such as precision, recall, F1-score, average precision (AP), and inference time, providing insights into how different models handle the complexities of coronary stenosis detection. Furthermore, techniques like Focal Loss are discussed, addressing class imbalance challenges and improving the detection of smaller objects. The analysis reveals that R-CNN models deliver higher detection accuracy ( \(86.34\%\) ), while the SSD architecture outperforms others in terms of real-time processing capabilities (0.2869 s average inference time). These findings underscore the potential of DL models to support healthcare diagnostics by enabling faster and more precise image analysis, facilitating early diagnosis, and enhancing clinical workflows.