Coronary Artery Stenosis and High-Risk Plaque Assessed With an Unsupervised Fully Automated Deep Learning Technique

Area of Science:

• Cardiology
• Artificial Intelligence
• Medical Imaging

Background:

• Coronary computed tomography angiography (CCTA) is vital for assessing coronary artery stenosis and high-risk plaque (HRP).
• Current CCTA analysis for stenosis and HRP is time-consuming and requires specialized expertise, hindering widespread clinical adoption.

Purpose of the Study:

• To develop and validate a fully automated deep learning system for characterizing stenosis severity and HRP on CCTA.
• To enhance the efficiency and accessibility of CCTA interpretation.

Main Methods:

• A deep learning system was trained on CCTA scans from 570 patients across multiple centers.
• The system assessed stenosis severity (>0%, 1-49%, ≥50%, ≥70%) and HRP features (low attenuation plaque, positive remodeling, spotty calcification).
• Model validation involved testing on 769 patients for stenosis and 45 patients for HRP.

Main Results:

• The deep learning system achieved 93.5% per-vessel agreement within 1 Coronary Artery Disease-Reporting and Data System (CAD-RADS) category for stenosis.
• Excellent diagnostic performance was observed for stenosis detection (>0% and ≥50%) and HRP characterization (AUCs ranging from 0.77 to 0.80).
• The system demonstrated high sensitivity, specificity, PPV, and NPV for stenosis assessment.

Conclusions:

• A fully automated, unsupervised deep learning system can rapidly evaluate stenosis severity and characterize HRP on CCTA.
• The system exhibits very good diagnostic performance, paving the way for improved clinical translation of CCTA.
• This AI-driven approach promises to streamline CCTA interpretation and support clinical decision-making.