Coronary Artery Stenosis and High-Risk Plaque Assessed With an Unsupervised Fully Automated Deep Learning Technique
Authors' conclusion
A fully automated unsupervised deep learning system can rapidly evaluate stenosis severity and characterize HRP with very good diagnostic performance on CCTA.
Study context
- Population
- Patients undergoing coronary CT angiography (coronary computed tomography angiography (CCTA))trained 570
- Comparison
- Fully automated deep learning system vs reference CCTA assessment
- Outcome
- Diagnostic performance for stenosis severity and high-risk plaque (HRP) on CCTA
Methodological detailsmulticentre
- Multicentre
Performance
As reported in the abstract
| Test or model | Per-vessel agreement within 1 Coronary Artery Disease-Reporting and Data System (CAD-RADS) category | >0% stenosis sensitivity | >0% stenosis specificity | >0% stenosis PPV | >0% stenosis NPV | >0% stenosis AUC | ≥50% stenosis sensitivity | ≥50% stenosis specificity | ≥50% stenosis PPV | ≥50% stenosis NPV | ≥50% stenosis AUC | Low attenuation plaque AUC | Spotty calcification AUC | Positive remodeling AUC |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Deep learning system | 93.5% | 90.6% | 88.8% | 83.4% | 93.9% | 89.7% | 87.1% | 92.3% | 60.9% | 98.1% | 89.7% | 0.80 | 0.79 | 0.77 |
Deep learning system
Per-vessel agreement within 1 Coronary Artery Disease-Reporting and Data System (CAD-RADS) category93.5%
>0% stenosis sensitivity90.6%
>0% stenosis specificity88.8%
>0% stenosis PPV83.4%
>0% stenosis NPV93.9%
>0% stenosis AUC89.7%
≥50% stenosis sensitivity87.1%
≥50% stenosis specificity92.3%
≥50% stenosis PPV60.9%
≥50% stenosis NPV98.1%
≥50% stenosis AUC89.7%
Low attenuation plaque AUC0.80
Spotty calcification AUC0.79
Positive remodeling AUC0.77
Original publication
Coronary Artery Stenosis and High-Risk Plaque Assessed With an Unsupervised Fully Automated Deep Learning Technique. JACC Adv. Published 6 March 2024. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →