Damini Dey PhD, Presenter: Nothing to Disclose

Ryo Nakazato, Abstract Co-Author: Nothing to Disclose

Victor Y Cheng, Abstract Co-Author: Nothing to Disclose

Piotr Slomka PhD, Abstract Co-Author: Nothing to Disclose

Guido Germano, Abstract Co-Author: Nothing to Disclose

Daniel S. Berman MD, Abstract Co-Author: Research grant, Lantheus Medical Imaging, Inc Research grant, Astellas Group Research grant, Siemens AG Speaker, Bristol-Myers Squibb Company Speaker, Covidien AG Speaker, Astellas Group Stockholder, Spectrum Dynamics Ltd Royalties, CSMC Consultant, Bracco Group Consultant, FlouroPharma, Inc

Purpose: We recently developed automated computer software (APQ) for three-dimensional (3D) coronary artery and plaque segmentation, and quantification of non-calcified and calcified plaque (NCP, CP) from coronary CT Angiography (CCTA). The aim of this study was to determine accuracy of APQ for 3D quantitative plaque analysis compared to expert readers.

Methods: APQ combines derivation of scan-specific attenuation thresholds for lumen, NCP and CP and knowledge-based segmentation and modeling of the coronary arteries to quantify plaque components in 3D. CCTA data from 27 consecutive patients with atherosclerotic plaques in proximal-mid coronary arteries, acquired with the dual-source CT scanner (Siemens Definition), were analyzed with APQ. Coronary artery Remodeling Index (RI) was defined as the ratio of vessel areas from the most stenosed cross-section with the normal proximal cross-section. The presence of positive remodeling (PR) was defined as RI>1.05. NCP and CP volumes, 3D plaque composition (defined by % of plaque volume classified as NCP and CP) and RI were computed using APQ, and were also manually quantified by 2 blinded expert readers for reference.

Results: Analysis time was <20 sec per lesion for APQ and 5-25 minutes for manual quantitation. There was excellent correlation between APQ and expert readers for NCP and CP volumes (r=0.93 and 0.88, p<0.0001), and for %NCP and %CP composition (r=0.87, and r=0.90, p<0.0001). Difference from expert readers for 3D plaque composition was 14.4±19% for APQ for NCP and 14.2 ± 18.3% for CP. For 21/27 non-calcified and mixed plaques, RI quantified by experts and APQ were 1.3±0.4 and 1.2±0.2 (p=0.08, NS), with excellent agreement between APQ and experts regarding PR presence (kappa 0.88, p=0.02).

Conclusions: Automated coronary artery segmentation and plaque quantification allows close agreement of 3D plaque volume and composition and coronary artery remodeling with expert readers, and requires a fraction of the time needed for manual analysis.

Automated coronary artery segmentation and plaque quantification allows close agreement of 3D quantitative plaque analysis compared to expert readers, at a fraction of time needed for manual analysis.

Dey, D, Nakazato, R, Cheng, V, Slomka, P, Germano, G, Berman, D, Automated Three Dimensional Quantitative Plaque Analysis from Coronary CT Angiography.  Radiological Society of North America 2010 Scientific Assembly and Annual Meeting, November 28 - December 3, 2010 ,Chicago IL. http://archive.rsna.org/2010/9013635.html