The Accuracy and Efficiency of AI in Coronary and Carotid-cerebral Atherosclerosis using One-step CTA.

INTRODUCTION

While AI has shown promise in single-site imaging, its utility in multi-vascular protocols remains underexplored. To our knowledge, this is the first study to evaluate the diagnostic efficiency and accuracy of an artificial intelligence (AI)-based computer-aided diagnosis (CAD) platform for evaluating one-step intra- and extracranial combined cardiovascular computed tomography angiography (ICCC-CTA).

METHODS

A total of 150 patients who underwent one-step coronary and carotid-cerebrovascular CTA examinations from April 2021 to June 2024 were retrospectively enrolled. Among these, 65 and 72 patients underwent carotid-cerebrovascular DSA and coronary DSA examinations, respectively. The one-step CTA images of all patients were evaluated independently by the CAD platform and senior and junior radiologists.

RESULTS

A reduction in CAD platform diagnostic time for coronary stenosis of 80.64% and 87.45%, and a reduction in mouse-click count of 41.94% and 57.77%, was observed compared to the senior and junior radiologist groups. The diagnostic time on the CAD platform for carotid-cerebrovascular stenosis was reduced by 79.92% and 88.33%, and the number of clicks by 24.24% and 42.76%, compared with the senior and junior radiologist groups. The CAD-assisted system demonstrated favorable diagnostic accuracy, yielding high area under the Curve (AUC) values for evaluating coronary and carotid-cerebral artery stenosis at the per-patient, per-vessel, and per-segment levels.

DISCUSSION: Previously reported literature has primarily focused on artificial intelligence systems using single-anatomical-site CT scans. In contrast, the present study utilizes an AI system to evaluate vascular stenosis using comprehensive, one-step coronary and carotid-cerebral CT images. Through this single-scan approach, the CAD platform demonstrates the capacity to automatically segment both the coronary and carotid-cerebral vasculatures.

CONCLUSION

Our findings demonstrate that the AI-powered CAD platform significantly improves both diagnostic efficiency and diagnostic accuracy relative to conventional, manually performed ICCC-CTA interpretation acquired via the one-step ICCC-CTA.