In the last decade, several MRI sequences have been developed for direct visualization of the intracranial vessel wall.¹ Their advantage, compared with MRA or CTA, is the ability to detect vessel wall lesions before any luminal stenosis is present. A vessel wall lesion is generally defined as a focal or more diffuse but restricted area of vessel wall thickening. However, vessel wall diseases, like atherosclerosis and vasculitis, are known to cause more generalized vessel wall pathology as well. Differentiating between these disease states and normal ageing, which has also been associated with non-pathological diffuse vessel wall thickening, is still not possible. Although much is known about vessel wall (intima-media) thickness of extracranial arteries², less is known about the intracranial arterial vessel wall. Furthermore, currently used MRI sequences have limited resolution and with in vivo imaging there is a lack of histological confirmation^3,4. In the current study, vessel wall thickness of major intracranial arteries was measured in ex vivo samples of the circle of Willis, using 7T MRI and histological validation, to ultimately provide a reference guide for normal intracranial vessel wall thickness.Fifteen anonymous circle of Willis (CoW) specimens were selected for this study. The specimens included the major arteries of the CoW. Before scanning, all specimens were cleaned from clotted blood products and embedded in a petri dish containing 2% agarose solution. Cactus spines were used as fiducials and placed at 15 locations to enable spatial correlation with histology⁵ (Figure 1). Imaging was performed on a 7T whole body system (Philips Healthcare), with a custom-made high density receive coil (16-channels per 70 cm²; MR Coils BV), and a volume transmit/receive coil for transmission (Nova Medical). For image acquisition, a T₁-weighted sequence was used with the following scan parameters: FOV 150x150x20mm, acquired resolution 0.13x0.13x0.13mm, TR/TE 26/4.3ms, flip angle 44 degrees, bandwidth 165 Hz/pixel, TFE factor 1154, acquisition time approximately 1h35min. After scanning, samples were taken from the 15 marked locations of each CoW specimen for histologic processing. Histologic processing was performed using an in-house developed protocol, as previously described⁵. The MR images were analyzed using the software program CAAS (Cardiovascular Angiographic Analysis System) MRA (Pie Medical). For initial validation of our technique, image reconstructions were made perpendicular to the relevant arteries at the fiducial locations, and matched to the corresponding histological section. The vessel wall boundaries (outer wall and lumen) were drawn manually on the histological sections (using MeVisLab; MeVis Medical Solutions AG) and the matching MRI slices (using CAAS MRA); an example is shown in Figure 2. Two methods were used for mean arterial wall thickness (MAWT) calculation⁶: (1) based on the distance between outer wall and lumen boundaries on multiple locations, and (2) based on the vessel areas. MR measurements were compared to histological measurements by linear regression. After initial validation, MR based MAWT calculations were performed for the major arterial segments of the CoW over the entire length of the specific segments.Initial validation results of the vessel wall thickness measurements of the first analyzed specimen are shown in Figure 3. Five samples were excluded due to lack of match with the corresponding MR image (n=1), or the histological section was too fragmented for thickness measurements (n=4), resulting in 10 samples that were used for analysis. The MAWT measured on MRI had a high goodness-of-fit with histology for both used methods, slightly favoring the method based on the vessel areas (R²=0.94 and R²=0.98, respectively). The root-mean-square-error (RMSE) was also lower for the vessel area method (RMSE=0.12 versus RMSE=0.07), however, the Bland-Altman plot indicates there is a diameter-dependent error for this method (Figure 3D). Figure 4 shows four examples of MAWT measurements of different arterial segments. Arterial segments had significantly varying MAWT, both between segments and within segments; for instance, wall thickness of the basilar artery varied from 0.43-1.03mm.The current results show that ultrahigh-resolution MRI at 7T enables accurate measurement of vessel wall thickness in ex vivo CoW specimens, with excellent goodness-of-fit correlations. Our results on vessel wall measurement of entire arterial segments so far have shown significant variation both within and between arterial segments. Using this method for more CoW specimens, we will be able to obtain reference values of intracranial vessel wall thickness variations. Not only will these values shed light on the spatial resolution actually required when visualizing intracranial vessel wall pathology in vivo, they can also ultimately become a reference standard when assessing intracranial vessel wall pathology, specifically when differentiating between pathological thickening and normal (age-related) variation.