Evaluation of Coronary Artery Image Quality with Knowledge-based Iterative Model Reconstruction

Abstract

To evaluate knowledge-based iterative model reconstruction (IMR) to improve image quality and reduce radiation dose in coronary computed tomography angiography (cCTA). We evaluated 45 consecutive cCTA studies, including 25 studies performed with an 80% systolic dose reduction using tube current modulation (TCM). Each study was reconstructed with filtered back projection (FBP), hybrid iterative reconstruction (iDose(4)), and IMR in a diastolic phase. Additional systolic phase reconstructions were obtained for TCM studies. Mean pixel attenuation value and standard deviation (SD) were measured in the left ventricle and left main coronary artery. Subjective scores were obtained by two independent reviewers on a 5-point scale for definitions of contours of small coronary arteries (<3 mm), coronary calcifications, noncalcified plaque, and overall diagnostic confidence for the presence/absence of stenosis. There was no significant difference in pixel intensity among FBP, iDose(4), and IMR (P > .8). For diastolic phase images, noise amplitude in the left main coronary artery was reduced by a factor of 1.3 from FBP to iDose(4) (SD = 99 vs. 74; P = .005) and by a factor of 2.6 from iDose(4) to IMR (SD = 74 vs. 28; P < .001). For systolic phase TCM images, noise amplitude in the left main coronary artery was reduced by a factor of 2.3 from FBP to iDose(4) (SD = 322 vs. 142; P < .001) and by a factor of 3.0 from iDose(4) to IMR (SD = 142 vs. 48; P < .001). All four subjective image quality scores were significantly better with IMR compared to iDose(4) and FBP (P < .001). The reduction in image noise amplitude and improvement in image quality scores were greatest among obese patients. IMR reduces intravascular noise on cCTA by 86%-88% compared to FBP, and improves image quality at radiation exposure levels 80% below our standard technique.