Standardized evaluation framework for evaluating coronary artery stenosis detection, stenosis quantification and...

Area of Science:

• Cardiovascular Imaging
• Medical Image Analysis
• Artificial Intelligence in Medicine

Background:

• Computed tomography angiography (CTA) is increasingly used for diagnosing coronary artery disease (CAD).
• Standardized evaluation of algorithms for CTA analysis is needed to ensure reliable clinical application.
• Conventional coronary angiography (CCA) remains a reference standard, but CTA offers a non-invasive alternative.

Purpose of the Study:

• To introduce a standardized framework for evaluating algorithms that detect and quantify coronary artery stenosis on CTA.
• To assess the feasibility of (semi-)automatic algorithms for coronary lumen segmentation in CTA.
• To compare algorithmic performance against quantitative coronary angiography (QCA) and expert annotations.

Main Methods:

• Development and application of a standardized evaluation framework for CTA algorithms.
• Quantitative evaluation of algorithms from 11 research groups using a database of 48 cardiac CTA datasets.
• Comparison of algorithmic stenosis detection/quantification with QCA and CTA consensus reading.
• Comparison of algorithmic lumen segmentation with expert manual annotation.

Main Results:

• Some current stenosis detection/quantification algorithms show potential for clinical triage or as a second reader.
• Automatic coronary lumen segmentation on CTA achieved precision comparable to expert manual annotation.
• The framework provides a reliable method for comparing the performance of different CTA analysis algorithms.

Conclusions:

• Dedicated algorithms demonstrate feasibility for detecting and quantifying coronary artery stenosis using CTA.
• Automatic lumen segmentation in CTA is a viable tool with expert-level precision.
• The developed framework facilitates the advancement and clinical integration of AI in cardiovascular imaging.