Intra- and interobserver variability in detection and assessment of calcified and noncalcified coronary artery plaques using 64-slice computed tomography

Objective Multislice computed tomography (MSCT) is an emerging noninvasive technique for detecting coronary plaques. The present study investigated agreement in the detection and characterization of coronary plaques and reproducibility of volumetric analysis. Methods A total of 20 patients underwent MSCT coronary angiography using 64 * 0.5 mm detector collimation. Two readers independently visually evaluated all MSCT datasets for the presence of coronary plaques (n = 82 in 262 coronary segments) and then classified them as calcified, mixed and noncalcified. In addition, one of the readers also manually determined total volumes as well as calcified and noncalcified volumes of each plaque. After a period of at least 4 weeks the complete volumetric analysis was repeated. Results Interobserver agreement was good for detection of coronary plaques on the segment level (weighted κ = 0.88, 95% CI [0.76, 0.95]). However, there was only moderate interobserver agreement for plaques classification (unweighted κ = 0.45, 95% CI [0.35, 0.61]). Intraobserver agreement was good for plaque detection on segment level (weighted κ = 0.90, 95% CI [0.77, 0.96]), while it was moderate with respect to their characterization (unweighted κ = 0.65, 95% CI [0.55, 0.80]). There was moderate reproducibility for total plaque volume (limits of agreement = ±6 mm³ at a mean measured volume of 10 mm³ and = ±28 mm³ at a mean measured volume of 100 mm³). Variation of relative differences significantly decreased for total volume and noncalcified volume with increasing mean volume. Conclusions Detection and volumetry of plaques by means of MSCT shows good to moderate reproducibility. Agreement of volume measurements depends on plaque size. Variation of relative differences decrease with increasing mean plaque volume.