This paper addresses the challenge of automatic detection of coronary stenosis with a new hybrid deep learning architecture. The proposed architecture combines the strengths of Single Shot MultiBox Detector (SSD) for efficient region proposal generation and Faster Region-based Convolutional Neural Network (Faster R-CNN) for accurate classification and bounding box refinement. By integrating these two architectures, SR-CNN optimizes both detection accuracy and computational efficiency, making it suitable for real-time medical applications. SR-CNN introduces a more efficient region proposal generation approach, reducing the number of candidate regions by approximately 75% compared to Faster R-CNN, leading to a significant decrease in computational cost. SR-CNN was compared against state-of-the-art object detection models, including Faster R-CNN, SSD, YOLOv8, YOLOv9, and RetinaNet. Experimental results demonstrate that SR-CNN achieves an F1-score of 0.8232, outperforming all other tested models, while maintaining a competitive inference time of 3.2901 s. In summary, SR-CNN contributes a novel hybrid framework that can enhance real-time clinical decision-making for coronary stenosis detection.