Artificial intelligence-based quantitative coronary angiography of major vessels using deep-learning

Area of Science:

• Cardiology
• Medical Imaging
• Artificial Intelligence

Background:

• Quantitative coronary angiography (QCA) provides objective coronary lesion assessment but suffers from observer variability and time constraints.
• Current clinical practice faces challenges with on-site QCA due to these limitations.

Purpose of the Study:

• To develop and evaluate a novel artificial intelligence-based quantitative coronary angiography (AI-QCA) method for analyzing major coronary vessels.
• To assess the performance of AI-QCA in comparison to manual QCA.

Main Methods:

• AI-QCA was developed using deep learning models trained on 7658 angiographic images for lumen boundary delineation.
• An automated quantification method with refined matching and iterative updates was integrated into AI-QCA.
• A retrospective analysis compared AI-QCA with manual QCA on 676 coronary angiography images, evaluating diameter stenosis (DS), minimum lumen diameter (MLD), reference lumen diameter (RLD), and lesion length (LL).

Main Results:

• AI-QCA demonstrated 89% sensitivity in lesion detection with strong correlations to manual QCA for DS, MLD, RLD, and LL.
• 80% of matched lesions (892/995) showed DS differences of ≤10% compared to manual QCA.
• Multiple coronary lesions were accurately identified and quantitatively analyzed by AI-QCA without manual intervention.

Conclusions:

• AI-QCA shows significant potential as an automated tool for coronary angiography analysis.
• This AI-driven approach may enhance quantitative assessment of coronary lesions and support clinical decision-making.