Twenty patients with a 0–29% stenosis grade (based on NASCET  criteria) and twelve patients for each of the three other stenosis grades (30–49%; 50–69% and 70–99%) at the symptomatic side were retrieved at random from a database (n = 421) of MDCTA examinations of patients with transient ischemic attack or minor stroke. In all 56 patients MDCTA had been performed as part of a research protocol that was approved by the Institutional Review Board and for which all patients had given written informed consent.
Scanning and image reconstruction
Scanning was performed on a 16-slice MDCT scanner (Siemens, Sensation 16, Erlangen, Germany) with a standardized protocol (120 kVp, 180 mAs, collimation 16 × 0.75 mm, table feed 12 mm/rotation, pitch 1) . All patients received 80 ml contrast material (320 mg/ml), followed by 40 ml saline, both with an injection rate of 4 ml/s . The radiation dose was 2.6 mSv.
Image reconstructions were made with field of view 120 mm, matrix size 512 × 512 (yielding interpolated pixels of 0.2 × 0.2 mm, real in-plane resolution is 0.6 × 0.6 mm), slice thickness 1.0 mm, increment 0.6 mm and with an intermediate reconstruction algorithm (B46: heart-view sharp) .
Quantification and characterization
Three observers independently assessed the presence of an atherosclerotic lesion, the length of the atherosclerotic lesion, the location of the bifurcation, lumen attenuation, and plaque volume and plaque component volumes. One of the observers assessed after 4 months for a second time the volumes in a subset of patients (half the population per stenosis degree, randomly chosen).
The criterion used for the presence of an atherosclerotic lesion was: the presence of a calcification and/or thickening of the vessel wall. The length of the atherosclerotic lesion was defined as the distance between the first (most proximal) image and the last (most distal) image on which the atherosclerotic lesion was present. The location of the bifurcation was defined as the first image with two separate lumina. Lumen attenuation was measured in the most proximal and distal image with atherosclerosis, and the mean lumen attenuation was calculated.
Plaque and plaque component areas were measured with a polymeasure plug-in developed by one of the co-authors (E.M.) for the freely available software package ImageJ (Rasband, National Institute of Mental Health, Bethesda, USA). This plug-in made it possible to draw manually regions of interest (ROI) in consecutive axial MDCT images and to automatically calculate the total number of pixels and the number of pixels of different Hounsfield value (HV) ranges within these ROI (Fig. 1). The ROI was placed over the outer vessel wall contour and therefore equals plaque area plus lumen area. The different HV ranges are considered to represent the different plaque components; calcification >130 HU, fibrous tissue 60–130 HU and lipid core <60 HU.
The cut-off value between calcifications and fibrous tissue was set at 130 HU; the value currently used for calcium scoring. The cut-off value between fibrous tissue and lipid core was set at 60 HU as assessed in previous studies [5, 6]. The cut-off value between atherosclerotic plaque and lumen was adjusted for each patient and based on the full-width-half-maximum principle (mean lumen attenuation plus mean fibrous tissue attenuation (≈88 HU) divided by two). To compensate for partial volume effects, related to a high lumen attenuation at the plaque-lumen border, the pixels around the lumen with a HV between 130 HU and the adjusted cut-off value were considered to be fibrous tissue. To assess the border between lumen and atherosclerotic plaque it was necessary to draw a second ROI close to the lumen in each image. Normally, the lumen area was then automatically differentiated from atherosclerotic plaque based on the adjusted cut-off value. But in those plaques in which calcifications bordered the lumen and the two dense structures merged with each other, lumen area and calcifications had to be separated by manual drawing.
The volumes were calculated as the product of the number of pixels, the pixel size and the increment.
Firstly, the difference between observers in the assessment of the presence of an atherosclerotic lesion was assessed. Hereafter, consensus on the presence of an atherosclerotic lesion was achieved by a consensus reading between all three observers. Those image series that were appointed as having atherosclerosis were used for further analysis.
Secondly, differences between observers in the assessment of the length of the atherosclerotic lesion, the location of the bifurcation, lumen attenuation and plaque and plaque component volumes, were calculated.
After assessment of the differences, a second consensus reading was held in order to achieve consensus about the length of the atherosclerotic lesion, the location of the bifurcation and lumen attenuation because these features influence the volume measurements. All observers had to adapt their assessments on grounds of this second consensus reading and hereafter plaque and plaque component volumes were calculated again and differences were evaluated. This recalculation provides observer variability measurements due to differences in the assessment of the outer vessel wall contour only.
In order to assess not only the variability in volume measurements, the overlap (similarity index) between the ROIs (outer contour) of the observers was assessed and expressed as a percentage (2 × pixels with overlap/(pixels ROI observer A + pixels ROI observer B)× 100%).
Finally, the intra-observer differences in plaque and plaque component volume measurements were assessed.
Continuous data were compared with a paired Student’s t-test for which a P-value < 0.05 was considered to indicate statistical significance.
Inter-observer differences in the assessment of the length of the atherosclerotic lesion, the location of the bifurcation and lumen attenuation were expressed as the mean ± the standard deviation (SD), and as a coefficient of variation defined by the SD of the paired difference divided by the mean of the absolute values.
Inter- and intra-observer differences in plaque and plaque component volume measurements were presented with a mean ± SD, an interclass correlation coefficient (ICC) with 95% confidence interval and a coefficient of variation. The differences were also plotted against the mean value of the measurements (Bland-Altman plot).