Vessel wall thickening and low wall shear stress are known as indicators of the development of atherosclerosis in the carotid bifurcation¹. The purpose of this study is to investigate if, already in early disease, a correlation exists between wall thickness (WT) and wall shear stress (WSS), both derived from MRI data, in the carotid bifurcation in an asymptomatic population. For this purpose, the individual relations between 3D WT and WSS maps are investigated, as well as the relation between the cohort-averaged 3D WT and WSS maps (as created with the atlas approach, as previously described for the aorta²). Furthermore, the atlases are downsampled to investigate the influence of resolution on the amount of spatial autocorrelation of the atlases³. Eleven subjects (mean age: 73±7 years, range: 55-78 years) with plaques in the left carotid arteries were selected from the Rotterdam study⁴. The MRI examination, performed on a 1.5 GE Scanner (GE Signa Excite II; GE Healthcare, Milwaukee, WI, USA) included non-gated (time-averaged) 3D flow MRI (spatial resolution: 0.7x0.7x1 mm³, TE/TR: 4.3ms/13ms, Venc: 60 cm/s) and Proton Density weighted-EPI (PDw-EPI, spatial resolution: 0.5x0.5x1.2 mm³, TE/TR: 24.3ms/12000ms). The 3D flow MRI images were corrected for background phase offsets by subtraction of the velocity in static tissue. The lumen and vessel wall were manually segmented from the PDw-EPI images using ITK-snap. The segmentation of the wall was converted to a mesh to enable 3D WT calculations (quantified as the shortest distance between a mesh point and the outer wall). Rigid co-registration of the PDw-EPI and 3D flow MRI images was performed in Elastix to ensure that the identical lumen segmentation was used for 3D WSS. WSS was quantified using the 3D flow MRI images with the method that was described previously⁵. A ‘shared’ geometry of the 11 carotid bifurcations was created using a combination of rigid registration and varying thresholds of overlap as previously described for aortas². Subsequently, the WSS and WT values for each subject were interpolated (nearest neighbor) onto the ‘shared geometry’ followed by averaging over the number of subjects, such that each WT and WSS value on the map is an average of the 11 subjects, yielding WT and WSS atlases. Spearman’s correlation coefficient ρ was determined both for subject-specific WT and WSS maps, followed by a Fisher z-transformation (normalizing ρ) to enable a student’s t-test to test if the z-values averaged over the subjects was significantly different from 0. The p-value was considered significant when lower than 0.05. ρ was also calculated for the full and downsampled WT and WSS atlases. Downsampling was performed with factors 2, 4, 8 and 16. Figure 1 displays the WSS and WT maps for the subjects with the highest and lowest ρ. The carotid bifurcation geometries included the common carotid artery (CCA), the external carotid artery (ECA) and the internal carotid artery (ICA). Note that the lowest ρ was the only one indicating a positive relation between WT and WSS. All other ρ’s were negative. All individual relationships were significant with p<0.001. The averaged ρ over all subjects was -0.28, which, after the Fisher z-transformation, was significantly different from 0 (p<0.001). In figure 2 the WSS and WT atlases are shown. Regions of high WT show good co-localization with regions of low WSS. The ρ for the atlases was -0.66. For all downsampled atlases, as displayed in figure 3, the regions of low WSS corresponded with regions of high WT. The change in ρ was minimal for all downsampled atlases (table 1). The co-localization of regions of low WSS and high WT confirms that wall thickening occurs in regions of low WSS in early disease. For both the individual and atlas statistical analyses, this relationship was significant. The increased ρ compared to the individual analysis shows that the atlas approach, even after downsampling, facilitates statistical power to such observations. Extension of our database with more healthy and diseased subjects would facilitate a comparison of the individual atherosclerotic patient with a cohort of healthy subjects, to quantify and visualize the severity of atherosclerotic disease.The atlas approach for WT and WSS mapping is feasible in the carotid bifurcation and robust for resolution-related spatial autocorrelation. Individual and atlas-based WT measurements correlated negatively with individual and atlas-based WSS measurements. The techniques as presented may prove useful for the diagnosis and monitoring of plaque development in patients with atherosclerosis.