Automated computer-aided stenosis detection at coronary CT angiography: initial experience

Area of Science:

• Cardiovascular Imaging
• Artificial Intelligence in Medicine
• Radiology

Background:

• Coronary artery stenosis detection is crucial for cardiovascular disease management.
• Coronary CT angiography (cCTA) is a key imaging modality.
• Automated analysis tools can potentially improve diagnostic efficiency.

Purpose of the Study:

• To evaluate the performance of a computer-aided algorithm for automated stenosis detection.
• To compare the algorithm's accuracy against quantitative coronary angiography (QCA).

Main Methods:

• 59 patients undergoing cCTA and QCA were analyzed.
• A software algorithm performed automated, unsupervised detection of coronary artery stenosis.
• Algorithm performance was assessed for detecting stenosis of 50% or more.

Main Results:

• The algorithm correctly identified 74% of stenoses (28 out of 38) of 50% or more.
• Per-vessel/per-patient analysis showed 74%/100% sensitivity and 83%/65% specificity.
• Negative predictive values were high at 94% (per-vessel) and 100% (per-patient).

Conclusions:

• The automated algorithm demonstrates relatively high accuracy for diagnosing significant coronary artery stenosis at cCTA.
• Its high negative predictive value may aid in confidently excluding stenosis.
• The algorithm could serve as a valuable second reader in cCTA interpretation.