High scan-rescan repeatability of AI-enabled coronary plaque quantification from coronary CT angiography

OBJECTIVES: To evaluate the reproducibility of plaque quantification from serial cardiac computed tomography (CCTA) scans by a systematic side-by-side approach using validated AI-enabled software with both scan-specific and fixed attenuation thresholds. MATERIALS AND METHODS: Thirty participants from two centers underwent serial CCTA within a short timeframe (median 6 days [IQR 0-30]). Volumes and burden (defined as plaque volume indexed by vessel volume) of total plaque (TP), calcified plaque (CP), non-calcified plaque (NCP), and low-density NCP (LD-NCP) were quantified per-patient while comparing the serial scans side-by-side. Analyses were done by two assessors using consensus readings to reduce individual bias, and inter-scan differences were compared using mean differences and the repeatability coefficient (RC, defined as 1.96 × standard deviation). RESULTS: ), and for burden 0.6%, 1.2%, and 0.3% for CP, NCP, and LD-NCP, respectively (RC: 1.3%, 2.4%, and 0.8%). Scan-specific thresholds presented with lower repeatability coefficients and narrower limits of agreement than fixed thresholds. CONCLUSION: Scan-rescan reproducibility was high across all plaque subcomponents for serial plaque quantification using a systematic assessment method and scan-specific attenuation thresholds with AI-enabled plaque software. KEY POINTS: Question Can comparisons of serial coronary CT angiography scans side-by-side offer reproducible quantifications of coronary plaque, and do scan-specific attenuation thresholds provide higher consistency than fixed thresholds? Findings AI-enabled plaque quantifications side-by-side were highly reproducible, and scan-specific attenuation thresholds provided higher consistency than fixed thresholds. Clinical relevance This study helps clinicians to reliably assess changes in coronary plaque volumes over time by defining the reproducibility across serial scans. The study also favors scan-specific attenuation thresholds over fixed thresholds for optimal reproducibility in the quantification of noncalcified plaque and calcified plaque.