Detection of coronary artery disease using faster R-CNN with X-ray angiography images

Coronary artery disease (CAD) is one of the leading causes of death globally, in which accurate detection and grading of stenosis is extremely important for effective treatment. While X-ray Coronary Angiography (XCA) is the diagnostic gold standard, its utility is hindered by drawbacks such as visual artifacts and limited annotated datasets. This study presents a Faster R-CNN-based deep learning method, a highly efficient object detection system, for stenosis detection and grading in coronary arteries from XCA images. The network is trained and tested on the ARCADE11 dataset, which is a large public expert-annotated XCA dataset. The architecture is customized with ResNet-50 and Feature Pyramid Networks for better feature learning. Our results exhibit high accuracy, low loss, and fast inference times, suggesting that the model is suitable for real-time clinical application. The model is nonetheless limited in real-time deployment by processing latency and is bound by the size of the training dataset, which may affect its capacity to generalize in real-world settings. Future work involves experimentation with a larger and more diverse dataset, experimenting with YOLO for improved detection, and improving performance metrics such as mAP, precision, and recall. Model optimization for real-time application is the goal, which will result in faster and more precise diagnosis in clinical practice.