Characteristics and Thresholds of Coronary Lesions to Predict Acute Coronary Syndrome From Artificial Intelligence-Guided Coronary Computed Tomography Angiography.
Authors
Affiliations (7)
Affiliations (7)
- Turku PET Centre, Turku University Hospital and University of Turku, Turku, Finland; Bern University Hospital Inselspital, Department of Cardiology, Bern, Switzerland. Electronic address: [email protected].
- Turku PET Centre, Turku University Hospital and University of Turku, Turku, Finland; Department of Clinical Physiology, Nuclear Medicine, and PET, Turku University Hospital, Turku, Finland.
- Turku PET Centre, Turku University Hospital and University of Turku, Turku, Finland; Heart Center, Turku University Hospital and University of Turku, Turku, Finland.
- Turku PET Centre, Turku University Hospital and University of Turku, Turku, Finland; Nuclear Medicine & PET, Department of Surgical Sciences, Uppsala University, Uppsala, Sweden.
- Heart Center, Turku University Hospital and University of Turku, Turku, Finland; Leiden University Medical Center, Department of Cardiology, Leiden, the Netherlands.
- Amsterdam University Medical Center, Department of Cardiology, Amsterdam, the Netherlands.
- Radboud University Medical Center, Department of Cardiology, Nijmegen, the Netherlands.
Abstract
Characteristics of future culprit lesions according to artificial intelligence-guided quantitative computed tomography (AI-QCT) are unknown. The authors aimed to characterize and derive thresholds to predict future culprit lesions from coronary computed tomography angiography using AI-QCT. A total of 2,169 patients/13,772 lesions undergoing coronary computed tomography angiography at Turku University Hospital, Finland, (derivation) and 553 patients/3,948 lesions from Amsterdam University Medical Center, the Netherlands, (validation) were analyzed with AI-QCT. Lesions from vessels undergoing early revascularization were excluded. Lesion-level total plaque volume (TPV) and noncalcified plaque volume (NCPV) thresholds were derived and externally validated to predict lesion-level and patient-level acute coronary syndrome (ACS). Future culprit lesions from both cohorts (n = 69 validation; n = 15 derivation) had consistently higher median TPV (101 mm<sup>3</sup> validation; 115 mm<sup>3</sup> derivation; P = 0.811) and NCPV (65 mm<sup>3</sup> validation; 74 mm<sup>3</sup> derivation; P = 0.709) compared with control lesions (P < 0.001) and were nonobstructive in 69.6% (derivation) and 40.0% (validation). Thresholds of TPV ≥55 mm<sup>3</sup> and NCPV ≥32 mm<sup>3</sup> were derived and associated with significantly higher adjusted rates of lesion-level (TPV ≥55 mm<sup>3</sup>: adjusted HR: 12.09; 95% CI: 6.95-21.04; P < 0.001; NCPV ≥32 mm<sup>3</sup>: adjusted HR: 10.82; 95% CI: 6.22-18.80; P < 0.001) and patient-level ACS (TPV ≥55 mm<sup>3</sup>: adjusted HR: 2.92; 95% CI: 1.65-5.16; P < 0.001; NCPV ≥32 mm<sup>3</sup>: adjusted HR: 5.11; 95% CI: 2.49-10.46; P < 0.001) with consistent results in the validation cohort. Future culprit lesions vs nonculprit lesions by AI-QCT demonstrated significantly higher plaque volumes and were nonobstructive in 40% to 70%. Thresholds of TPV ≥55 mm<sup>3</sup> and NCPV ≥32 mm<sup>3</sup> were derived to predict future culprit lesions and associated with increased rates of lesion-level and patient-level ACS in 2 independent cohorts.