AI-Enhanced Coronary Artery Disease Diagnostics from X-Ray Angiography
This project develops AI-assisted tools to support cardiologists in accurately segmenting coronary arteries and detecting stenosis from X-ray angiography images, improving diagnostic consistency and efficiency in resource-limited settings and establishing foundations for diverse population deployment.
Coronary Artery Disease (CAD) is a leading cause of death and disability worldwide, causing 17.8 million deaths annually. Early and accurate diagnosis using coronary angiography is essential but limited in low-resource settings like Nepal due to a shortage of trained cardiologists. Manual analysis is time-consuming and subject to variability, creating a need for automated, reliable diagnostic support.
Our goal is to enhance CAD diagnostics in low-resource settings by developing AI tools that assist cardiologists with accurate artery segmentation and stenosis detection. The research focuses on building models that perform reliably across diverse patient populations and varied clinical environments.
The team has developed and benchmarked AI models using YOLOv8 with a pseudo-label-based data augmentation pipeline, improving segmentation performance and consistency. The method has increased F1 scores by 9% on validation and 3% on test datasets, demonstrating more accurate and reliable artery segmentation. These models provide a foundation for supporting cardiologists in clinical decision-making, with the potential to deploy AI-assisted CAD diagnostics widely in low-resource settings.