Montreal Heart Institute (HeartWise.AI) (Avram et al.)
AI-driven pipeline for quantitative coronary-stenosis assessment from routine DICOM coronary angiography videos, combining vessel tracking with a video Swin3D transformer trained and validated on 182,418 angiography videos spanning 5 years at the Montreal Heart Institute. Achieves a mean absolute error of 20.15% and a classification AUROC of 0.8294 for stenosis-percentage prediction against cardiologist assessment, with lower inter-rater variability than two expert interventional cardiologists, and can be fine-tuned to quantitative coronary angiography (QCA) data for even lower error (MAE 7.75%).
Architecture
Hybrid
Multi-stage pipeline: Xception CNN for coronary segment/view identification, RetinaNet for stenosis-region detection, and a video Swin3D transformer for stenosis-percentage regression from coronary angiography video clips
Framework
PyTorch
Added to catalog
2026-08-10
Montreal Heart Institute Coronary Angiography Archive (DeepCoro)
182,418 coronary angiography videos spanning 5 years at the Montreal Heart Institute, used to train and validate DeepCoro for coronary segment identification and stenosis-percentage prediction.
Coronary artery stenosis percentage prediction from angiography video (Swin3D regression + severity classification)
Coronary artery segment (anatomical location) identification from angiography video vessel tracking