Abstract 13665: Determination of Lipid-Rich Plaque by Artificial Intelligence-Enabled Quantitative Computed Tomography Using Near-Infrared Spectroscopy

Abstract

Introduction: By near-infrared spectroscopy (NIRS), plaques with 4-mm maximum lipid-core burden index (maxLCBI4 mm ) ≥400 are lipid-rich plaques. By CCTA, high-risk plaques are defined by presence of low-density non-calcified plaque (LD-NCP) and positive remodeling (PR) >1.1. Hypothesis: Artificial intelligence-enabled quantitative CT (AI-QCT) will enable identification of lipid-rich plaques when compared to NIRS reference standard. Methods: 128 atherosclerotic plaques from 47 patients were prospectively enrolled. MaxLCBI 4mm (cutoff was defined as 400) and LD-NCP derived from AI-QCT (Cleerly LABS, Denver CO) were compared. We also analyzed vessel area, plaque burden, minimal lumen area and plaque length between intravascular ultrasound (IVUS) and AI-QCT. Results: Amongst 128 plaques evaluated, maxLCBI 4mm ≥400 was 26 plaques (20.3%). Use of current LD-NCP (<30 Hounsfield threshold (HU) and LD-NCP volume >0 mm 3 ) and PR ≥1.1, yielded a sensitivity, specificity, positive and negative predictive value and accuracy of 100%, 28%, 26%, 99%, and 43%. LD-NCP volume threshold was identified from AUC as 1.8 mm 3 . With LD-NCP volume ≥1.8 mm 3 , the performance of AI-QCT was improved to 92%, 87%, 65%, 98% and 88%, respectively (Figure 1). We noted very good to excellent correlation between AI-QCT and IVUS for vessel area (R2 =0.776), plaque burden (0.699), minimal lumen area (0.815) and plaque length (0.844). Conclusions: The current CCTA definition of NIRS-verified lipid-rich plaques has limited specificity and accuracy; use of a LD-NCP volume threshold, enabled only by AI-QCT, substantially improves performance for detection and characterization.