Automated 3-dimensional quantification of noncalcified and calcified coronary plaque from coronary CT angiography

Abstract

Introduction: We aimed to develop an automated algorithm (APQ) for accurate volumetric quantification of non-calcified (NCP) and calcified plaque (CP) from coronary CT angiography (CCTA).

Methods: APQ determines scan-specific attenuation thresholds for lumen, NCP, CP and epicardial fat, and applies knowledge-based segmentation and modeling of coronary arteries, to define NCP and CP components in 3D. We tested APQ in 29 plaques for 24 consecutive scans, acquired with dual-source CT scanner. APQ results were compared to volumes obtained by manual slice-by-slice NCP/CP definition and by interactive adjustment of plaque thresholds (ITA) by 2 independent experts.

Results: APQ analysis time was <2 sec per lesion. There was strong correlation between the 2 readers for manual quantification (r = 0.99, p < 0.0001 for NCP; r = 0.85, p < 0.0001 for CP). The mean HU determined by APQ was 419 +/- 78 for luminal contrast at mid-lesion, 227 +/- 40 for NCP upper threshold, and 511 +/- 80 for the CP lower threshold. APQ showed a significantly lower absolute difference (26.7 mm(3) vs. 42.1 mm(3), p = 0.01), lower bias than ITA (32.6 mm(3) vs 64.4 mm(3), p = 0.01) for NCP. There was strong correlation between APQ and readers (R = 0.94, p < 0.0001 for NCP volumes; R = 0.88, p < 0.0001, for CP volumes; R = 0.90, p < 0.0001 for NCP and CP composition).

Conclusions: We developed a fast automated algorithm for quantification of NCP and CP from CCTA, which is in close agreement with expert manual quantification.