Automatic assessment and blood flow distribution of the major cerebral arteries have applications in various brain diseases¹, for example in acute stroke care. The cerebral arterial system is complex with large inter-individual spatial variations, which are potentially challenging for automatic methods. Recently a method for automatic identification of cerebral arteries, that utilized a small sample probabilistic atlas, was presented². However, to reach its full potential this method will require a population based probabilistic cerebrovascular atlas. The objective of this work was to construct an artery specific probabilistic atlas of 16 large cerebral arteries, based on 168 subjects, and to investigate if the regional specificity of vascular branches was sufficient to permit atlas based arterial labeling. The atlas was based on 4D flow MRI³ from 168 subjects (65.6 ± 1.4 year old, 94 men, 74 female) collected on a 3 Tesla General Electric HD750 system and a 32-channel headcoil. FOV: 220 x 220 x 220 mm, voxel size: 0.7x0.7x0.7 mm3, scan time. 9.5 minutes, velocity encoding, 110 cm/s; TR/TE, 6.5/2.7 ms; flip angle, 8°. A temporal maximum intensity projection (tMIP) was reconstructed and used for arterial assessment. The tMIP for each subject was transformed to MNI-space using SPM8’s DARTEL⁴. The tMIP was smoothed with a low-pass-filter and binarized by thresholding at 18% of the maximum intensity value. The following arteries were segmented and labeled using an inhouse matlab tool: a. internal carotid arteries (ICA); b. vertebral arteries (VA); c. posterior cerebral arteries (PCA); d. proximal middle cerebral arteries (M1 segment); e. distal middle cerebral arteries (at M2-M3); f. proximal anterior cerebral arteries (A1 segment); g. distal anterior cerebral artery (at A2-A3); h. posterior communicating arteries (PCoA); i. basilar artery (BA). A probability map was constructed for each artery by averaging the binary volumes across subjects, these probability maps together formed the atlas. To investigate if the regional specificity of vascular branches was sufficient to permit atlas based arterial labeling, an internal analysis was performed. During arterial segmentation, a single-voxel-thick centerline (CL) was calculated for each artery by gradually thinning the binary image⁵. The average probability value along each CL was calculated, as well as the proportion of voxels in each CL that was assigned to the correct artery based on the highest probability value at its location. The rationale was that if >50% of the CL voxels were assigned to the correct artery, the arterial segment would be correctly labeled.The prevalence of arteries in the population based sample that was used for the probabilistic atlas is presented in Table 1. The highest mean probability value along the CL was obtained for ICA. High values were also seen for BA, M1, and A1, although there was a large variation among subjects (Table 1). The rate of correctly labeled CL-voxels was over 80% for all arteries, and over 90% for all arteries except PCoA. The main result of this study was the high accuracy of CL-voxel assignment supporting that the cerebrovascular arterial system permits atlas based arterial labeling. The average probability along the CL is related to the spatial variation across subjects. The closer to each other the arteries are located, the higher the average probability value. However, a lower mean probability along the centerlines did not equal a low rate of correct labeling. For example, the lowest probability was obtained for M2-M3, but the corresponding rate of correct labeling was among the highest because it has a more isolated spatial territory. We successfully constructed an artery specific probabilistic atlas that included 16 major cerebral arteries. The atlas showed a regional specificity of vascular branches that was sufficient to permit atlas based arterial labeling. This is a first step toward medical applications such as automatic arterial vessel identification, improved vascular segmentation, automatic blood flow assessment, and characterization of cerebral arterial morphology.