Impact of advanced modeled iterative reconstruction on interreader agreement in coronary artery measurements

Abstract

ObjectivesTo evaluate the influence of advanced modeled iterative reconstruction (ADMIRE) on coronary artery computed tomography angiography (cCTA) measurements in comparison to filtered back projection (FBP). Material and methodsPhantom scans and coronary CTA studies of 27 patients were acquired with a third generation dual-source CT scanner. Images were reconstructed using FBP and ADMIRE. Phantom measurements were used as reference standard. In patient studies, representative axial slices of each coronary artery segment without (n=308) and with coronary plaques (n=40) were assessed in identical positions for comparison of FBP and ADMIRE reconstructions. Image analyses included quality assessment, phantom and coronary artery measurements, plaque analysis, and interreader agreement of two independent and blinded readers. ResultsMean image noise was lower on ADMIRE reconstructions with 31.3±9.9 HU compared to 55.9±15.7 HU on FBP reconstructions (p<0.001). Measurement precision and interreader agreement of both observers were assessed satisfactorily on phantom images in comparison to the full width half maximum method. In patients, correlation of lumen diameters of both observers improved using ADMIRE with a Pearson’s r=0.987 (95% confidence interval [CI], 0.983–0.989; p<0.001) compared to FBP images with r=0.939 (95% CI, 0.924–0.951; p<0.001). Applying ADMIRE, agreement of both observers for lumen diameter measurements significantly increased (p<0.001). This was also observed for the degree of stenosis (p<0.001) with r=0.560 using FBP (95% CI, 0.301–0.742) and with r=0.818 using ADMIRE (95% CI, 0.680–0.900). Plaque density measurements correlated closely with a Pearson’s r of 0.951 in FBP (95% CI, 0.909–0.974) and 0.967 in ADMIRE (95% CI, 0.939–0.983). ConclusionsAdvanced modeled iterative reconstruction significantly improves coronary artery assessment in coronary CTA in comparison to FBP by improved image quality due to image noise removal. This renders improved interobserver agreement for coronary lumen diameter and degree of stenosis measurements without influencing mean plaque attenuation.