A Pipeline for Collecting and Preprocessing Coronary Angiography Images to Enable Automated AI-Based SYNTAX Scoring

This project addresses the challenge of accurate coronary heart disease (CHD) diagnosis from X-ray coronary angiograms, where manual interpretation can suffer from observer variability and limited reproducibility. The study aims to support machine-learning and data-mining research by providing an annotated dataset for automated angiogram interpretation and coronary stenosis estimation. The dataset contains 231 coronary angiography X-ray images from 231 patients retrospectively collected from Shahid Madani Hospital, along with lesion-level angiographic metadata and SYNTAX score information. Source images were obtained from PACS/HIS archives in DICOM format and converted to JPG frames for processing. The researchers also describe a machine-learning-based frame selection method using MSSIM/SSIM to identify the most informative frames from angiography sequences. The SYNTAX score, calculated by a cardiologist using standard angiographic criteria, is used to stratify patients into risk groups and provide clinically meaningful labels for machine-learning tasks. Overall, the project’s main contribution is a curated dataset that integrates imaging and clinical annotations to enable the development of AI methods for CHD diagnosis, severity assessment, and risk classification.