Reader and platform reproducibility for quantitative assessment of carotid atherosclerotic plaque using 1.5T Siemens, Philips, and General Electric scanners

Abstract

To evaluate the platform and reader reproducibility of quantitative carotid plaque measurements. A total of 32 individuals with >or=15% carotid stenosis by duplex ultrasound were each imaged once by a 1.5T General Electric (GE) whole body scanner and twice by either a 1.5T Philips scanner or a 1.5T Siemens scanner. A standardized multisequence protocol and identical phased-array carotid coils were used. Expert readers, blinded to subject information, scanner type, and time point, measured the lumen, wall, and total vessel areas and determined the modified American Heart Association lesion type (AHA-LT) on the cross-sectional images. AHA-LT was consistently identified across the same (kappa = 0.75) and different scan platforms (kappa = 0.75). Furthermore, scan-rescan coefficients of variation (CV) of wall area measurements on Siemens and Philips scanners ranged from 6.3% to 7.5%. However, wall area measurements differed between Philips and GE (P = 0.003) and between Siemens and GE (P = 0.05). In general, intrareader reproducibility was higher than interreader reproducibility for AHA-LT identification as well as for quantitative measurements. All three scanners produced images that allowed AHA-LT to be consistently identified. Reproducibility of quantitative measurements by Siemens and Philips scanners were comparable to previous studies using 1.5T GE scanners. However, bias was introduced with each scanner and the use of different readers substantially increased variability. We therefore recommend using the same platform and the same reader for scans of individual subjects undergoing serial assessment of carotid atherosclerosis.