Artificial intelligence-based coronary computed tomography angiography quantification of atherosclerosis burden: comparison with intravascular ultrasound in the INVICTUS Registry.
Authors
Affiliations (26)
Affiliations (26)
- Department of Cardiovascular Medicine, Toho University Graduate School of Medicine, Toho University Omori Medical Center, Tokyo, Japan.
- Division of Cardiovascular Medicine, Department of Internal Medicine, Toho University Faculty of Medicine, Toho University Omori Medical Center, Tokyo, Japan.
- Department of Cardiovascular Medicine, Gifu Heart Center, Gifu, Japan.
- Sapporo Cardiovascular Clinic, Hokkaido, Japan.
- Edogawa Hospital Tokyo, Tokyo, Japan.
- Department of Cardiovascular Biology and Medicine, Juntendo University, Graduate School of Medicine, Tokyo, Japan.
- Tenyoukai Central Hospital, Kagoshima, Japan.
- Department of Cardiovascular Medicine, Dentistry and Pharmaceutical Sciences, Okayama University Graduate School of Medicine, Okayama, Japan.
- Yotsuba Circulation Clinic, Matsuyama, Japan.
- Department of Cardiovascular Medicine, Hiroshima University Graduate School of Biomedical and Health Sciences, Hiroshima, Japan.
- Department of General Internal Medicine 3, Kawasaki Medical School General Medical Center, Okayama, Japan.
- Okayama Red-Cross Hospital, Okayama, Japan.
- Department of Cardiology, Sakakibara Heart Institute, Tokyo, Japan.
- Department of Cardiology, Kindai University Faculty of Medicine, Osaka, Japan.
- Department of Cardiology, Tokai University, School of Medicine, Isehara-shi, Japan.
- Department of Cardiology, Toho University Sakura Medical Center, Chiba, Japan.
- Department of Cardiology, NHO Shikoku Cancer Center, Matsuyama, Japan.
- Department of Radiology, Ehime University Graduate School of Medicine, Matsuyama, Japan.
- Department of Cardiovascular Medicine, Toho University Ohashi Medical Center, Tokyo, Japan.
- Section of Cardiovascular Medicine, Department of Internal Medicine, Yale School of Medicine, New Haven, CT, USA.
- Yale Cardiovascular Research Group, Yale School of Medicine, New Haven, CT, USA.
- Department of Radiology, University of California, San Diego, La Jolla, CA, USA.
- Cardiovascular Research Foundation, New York, NY, USA.
- George Washington University School of Medicine and Health Sciences, Washington, DC, USA.
- Section of Cardiovascular Medicine, Department of Internal Medicine, Yale School of Medicine, New Haven, CT, USA. [email protected].
- Yale Cardiovascular Research Group, Yale School of Medicine, New Haven, CT, USA. [email protected].
Abstract
Automated artificial intelligence (AI)-based assessment of atherosclerosis burden applied to coronary computed tomography angiography (CCTA) can optimize image processing times, standardize interpretation, and minimize inter-observer variability. We investigated the diagnostic utility of AI-based CCTA quantification (AI-QCT) of coronary atherosclerosis in coronary segments co-registered with intravascular ultrasound (IVUS) of diseased and non-diseased segments. Patients who underwent CCTA and IVUS in the INVICTUS registry (ClinicalTrials.gov: NCT04066062) were enrolled. Images were analyzed by independent core laboratories blinded to each modality's findings. Vessel external elastic membrane (EEM), lumen, plaque volumes, plaque burden, and percent atheroma volume (PAV) were quantified in whole co-registered segments and subsegments containing non-calcified and low-attenuation plaques. A calcium index was calculated for the whole co-registered segment. A total of 108 vessels from 85 patients were included. Pearson's correlation demonstrated strong associations between AI-QCT and IVUS in quantifying the EEM volume (r = 0.899), lumen volume (r = 0.943), and plaque volume (r = 0.833), length-normalized PAV (r = 0.851), and calcium index (r = 0.960) in the whole-segment analysis. Strong correlations were seen for vessel, lumen, and plaque volumes in non-calcified (Pearson's coefficient: 0.95, 0.97, and 0.83, respectively) and low-attenuation (Pearson's coefficient: 0.90, 0.86, and 0.86, respectively) plaque segments. The minimum lumen area was 0.61 ± 1.18 mm<sup>2</sup> (95% CI, -0.83 to -0.38) smaller by AI-QCT than IVUS, with a similar lumen area stenosis (mean difference, 1.26 ± 24.17; 95% CI, -3.37 to 5.90). AI-QCT quantification of atherosclerosis burden showed high correlations and close agreement with IVUS in whole-segment and segments with non-calcified and low-attenuation plaques. Question Coronary atheroma burden is a powerful predictor of cardiovascular events. Can AI-based coronary CT angiography (CCTA) accurately quantify atherosclerotic burden across the full disease spectrum when compared with intravascular ultrasound (IVUS)? Findings AI-based CCTA quantification (AI-QCT) showed strong correlations with IVUS for plaque volume, burden, and calcium across whole coronary segments, including non-calcified and low-attenuation plaques. Clinical relevance AI-QCT provides rapid, automatic, and accurate atherosclerosis quantification without reader-dependent variability, enabling standardized cardiovascular risk assessment, treatment monitoring, and therapeutic decision-making across all disease severity spectrum in routine clinical practice.