Artificial intelligence-guided quantitative coronary CT angiography (AI-QCT) automated detection and occlusion length estimation of chronic total occlusions.
Authors
Affiliations (5)
Affiliations (5)
- Minneapolis Heart Institute, Abbott Northwestern Hospital, Minneapolis, MN, USA.
- Minneapolis Heart Institute, Abbott Northwestern Hospital, Minneapolis, MN, USA; Center for Coronary Artery Disease, Minneapolis Heart Institute Foundation, Minneapolis, MN, USA.
- Cardiovascular Imaging Research Center and Core Lab, Minneapolis Heart Institute Foundation, Minneapolis, MN, USA.
- Division of Cardiology, Department of Medicine, University of Washington, Seattle, WA, USA.
- Minneapolis Heart Institute, Abbott Northwestern Hospital, Minneapolis, MN, USA; Center for Coronary Artery Disease, Minneapolis Heart Institute Foundation, Minneapolis, MN, USA. Electronic address: [email protected].
Abstract
Artificial intelligence (AI)-enhanced coronary computed tomography angiography (CCTA) analyses may enhance the detection of chronic coronary total occlusions (CTOs) and facilitate pre-procedural planning for CTO percutaneous coronary intervention (PCI). Observational study of 50 consecutive patients enrolled in the PROGRESS-CTO registry with pre-procedural CCTA and AI-based quantitative computed tomography (AI-QCT). We evaluated the diagnostic accuracy of AI-QCT compared with advanced cardiac imagers for CTO detection on CCTA, and AI-QCT compared with visually assessed invasive angiography for CTO length estimation. Pre-procedural CCTA with AI-QCT analysis was performed for 50 consecutive CTO PCIs (82 % of patients were men, mean age 66.5 ± 10.5 years). The right coronary artery was the most commonly treated vessel (46 %). As compared to advanced cardiac imagers who identified CTO lesions in 40 patients (80 %), AI-QCT-based automated CTO detection resulted in a comparable detection rate and identified CTO lesions in 41 patients (82 %). In 9 cases (18 %), AI-QCT and imaging cardiologists disagreed: AI-QCT identified CTOs in 5 cases where imaging cardiologists did not, and imaging cardiologists identified CTOs in 4 cases where AI-QCT did not. AI-QCT CTO length measurements had moderate correlation with angiography-based measurements (r = 0.69; p < 0.001), with a mean difference of 0.27 ± 14.9 mm. The antegrade approach was the most common successful crossing strategy (48 %), and technical success was achieved in 86 % of cases. In patients undergoing CCTA, AI-QCT facilitates the automated detection of CTO lesions and enables estimation of the occlusion length which may enhance treatment planning of CTO PCI.