Coronary computed tomography angiography analysis using artificial intelligence for stenosis quantification and stent...

Area of Science:

• Cardiovascular Imaging
• Artificial Intelligence in Medicine
• Medical Diagnostics

Background:

• Coronary computed tomography angiography (CCTA) interpretation requires expertise to accurately assess stenosis severity.
• Inexperienced readers may overestimate coronary artery stenosis.
• Artificial intelligence (AI) offers potential auxiliary diagnostic tools for CCTA.

Purpose of the Study:

• To externally validate an AI-assisted analysis system for CCTA.
• To assess the AI system's capability in rapidly evaluating stenosis severity.
• To explore the integration of AI into routine CCTA clinical workflows.

Main Methods:

• A multicenter study with internal and external CCTA cohorts (April 2017-February 2023).
• Stenosis severity evaluated using Coronary Artery Disease Reporting and Data System (CAD-RADS) scores.
• AI system (InferRead CT Heart v1.6) used for stenosis quantification and automatic stent segmentation.
• Performance metrics included sensitivity, specificity, PPV, NPV, and Dice similarity coefficient (DSC).

Main Results:

• For ≥50% stenosis, external dataset performance: 88.0% sensitivity, 94.5% specificity, 90.0% PPV, 93.4% NPV.
• For ≥70% stenosis, external dataset performance: 91.9% sensitivity, 97.3% specificity, 85.0% PPV, 98.6% NPV.
• Cohen kappa for CAD-RADS categorization ranged from 0.72 to 0.81.
• AI stent segmentation achieved a DSC of 0.96±0.06.

Conclusions:

• The AI-assisted analysis system demonstrated high proficiency in CCTA stenosis quantification and stent segmentation.
• AI holds significant potential for advancing CCTA postprocessing techniques.
• The validated AI system could support clinical decision-making in CCTA interpretation.