Related Experiment Video
Updated: Jun 3, 2025
06:57
Author Spotlight: Advancing Cardiovascular Imaging - Introducing the Spatially Weighted Calcium Score for Early Disease Detection
Published on: September 22, 2023
925
Automated stenosis estimation of coronary angiographies using end-to-end learning
Christian Kim Eschen1, Karina Banasik1, Anders Bjorholm Dahl2
1Novo Nordisk Foundation Center for Protein Research, Faculty of Health and Medical Sciences, University of Copenhagen, Copenhagen, Denmark.
The International Journal of Cardiovascular Imaging
|January 9, 2025
Summary
Deep learning models accurately assess coronary artery stenosis from angiography, outperforming visual estimation. This AI approach offers a faster, more precise method for evaluating significant stenosis in clinical practice.
Area of Science:
- Cardiology
- Artificial Intelligence
- Medical Imaging
Background:
- Coronary angiography visual assessment for stenosis lacks accuracy.
- Fractional flow reserve and quantitative coronary angiography are time-consuming and costly.
- Deep learning offers potential for faster, more accurate stenosis evaluation.
Purpose of the Study:
- Develop and validate deep learning models for coronary artery stenosis assessment.
- Classify coronary artery cine loops (left/right) and identify stenosis.
- Compare deep learning model performance against established methods.
Main Methods:
- Developed deep learning models to classify cine loops and estimate stenosis.
- Utilized a large dataset of 19,414 patients and 332,582 cine loops.
- Validated models on internal (5056 patients) and external (608 patients) test sets.
Main Results:
- Internal testing showed high accuracy (ROC-AUC 0.903) for significant stenosis detection.
- External testing demonstrated superior performance over visual assessment (ROC-AUC 0.833).
- Model performance was also strong against 3D QCA (0.798) and FFR (0.780).
Conclusions:
- Deep learning models show significant promise for accurate coronary stenosis prediction.
- This AI approach improves upon previous methods by including all segments and revascularized patients.
- The models offer a simpler, externally validated tool for stenosis assessment.
