The diameter of cerebral vessels is dynamic and thought to be mediated by neural and astrocytes signaling¹. While several studies in animal models have reported stimulus-induced vasodilation, measuring this in humans is not straightforward², although this could provide important information on neurovascular coupling. Here, we report on a novel and non-invasive MRI approach for quickly assessing changes in cerebral arterial diameter, which is then used to study the arterial response of the visual cortex, perfused by the posterior cerebral arteries (PCA³)

All MRI images were acquired on 4 healthy subjects at the Sherbrooke University Hospital Center (CHUS) using an Ingenia 3T Philips scanner. A low-resolution, 4-chunk, full brain time-of-flight (ToF⁴) (200 slices, TR/TE/Flip angle=23ms/3.6ms/18^o, 512x512x200 reconstructed voxels, acquisition resolution of 0.5x0.5x1.2mm, reconstructed resolution of 0.5x0.5x0.6mm, 2min acquisition time) was acquired as a scout to identify the PCA. Once the location of these arteries was determined, two high-resolution ToF with only one chunk (100 slices, TR/TE/Flip angle=23ms/3.6ms/18^o, 0.3mm isotropic reconstructed resolution, 5m45s acquisition time) were acquired in the same region: we first acquired a ‘resting state’ ToF where the subject had their eyes closed with the lights turned off. A second ToF was acquired while the subject viewed an action movie containing several scene cuts and contrast changes (no sound); both known to reliably activate visual neurons⁵. Compared to typical visual stimuli (i.e. checkerboards), this is advantageous to minimize subject fatigue given the relatively long stimulation period.

Pre-processing consisted of a registration of the two ToF followed by skull stripping with FSL’s BET and denoising by a non-local means via Dipy. For vessel segmentation, our method is based on Hessian-based vesselness measure⁶. Briefly, bright tubular structures are identified, segmented and stored for further analysis. We used vessel enhancing diffusion (VED⁷) to smooth the vessels and removes noise from the segmentation. To estimate diameter, a local thickness technique⁸ was performed by fitting the largest sphere in the segmentation binary mask. This resulted in a voxel-based diameter measure along each segmented vessel. Once segmented, the PCA was manually classified into their corresponding known sections (Basilar, P1, P2, P3, P4) by a radiology resident, which were then classified into five different clusters of similar sizes. A two sample paired t-test was used to assess significant differences in diameter between resting and activated states. For that, we computed the percentage change $$$(D-D_0)/D_0$$$ , where $$$D$$$ is the diameter of an active voxel and $$$D_0$$$ is the diameter of the same voxel in the rest image.

An example of a resting and active ToF maximum intensity projection (MIP) from one subject is shown in Figure 1. Clearly, stimulation enhanced the visibility of small vessels near the posterior portion of the occipital lobe; it can be seen by counting the number of new voxels in the active segmented image (~10.62±4.45%). A 3D reconstruction of the PCA, color-coded for diameter, along with the cluster classification used for computation is shown in Figure 2. Average apparent diameter values for all five PCA regions in rest and stimulus conditions are shown in Figure 3 for each participant (N=4).

In each subject (N=4), with stimulus-evoked a change in caliber was observed everywhere⁹, but there is a greater vasodilation in small vessels. This is evident when pooling data across all subjects and voxels spanning P4 and binning baseline diameter values in 10% increments (Figure 4). On average, stimulation dilated vessels in the 0.8mm range by ~20%, those around 1.6mm remained unchanged and vessels larger than 1.8mm appeared to constrict. In each subject, we found that around 57.44±16.04% of voxels dilate, 31.75±11.81% remain invariant and 10.81±6.25% constrict (Figure 5).

In summary, our data indicate that visual stimulation increases the apparent arterial diameter in the majority of the PCA. Such vasodilation is particularly strong in the smaller vessels near the cortex, in line with findings observed in animal models¹⁰. However, with the present data, we cannot rule out that the observed vasodilation results from enhanced image contrast (i.e. due to increased blood flow) rather than a true increase in diameter and we must take into account the partial volume effect, some vessels are not larger than the image resolution. We are currently acquiring measures of cerebral blood flow (CBF) to investigate this, as well as stimulus-evoked susceptibility weighted with phase difference (SWIp¹¹) acquisitions to confirm that venous dilation is indeed smaller than arterial dilation, as suggested in previous animal model studies¹⁰. Finally, we are currently experimenting with faster ToF acquisitions in order to measure the dynamic changes in vessel diameter on shorter time scales.