Cerebral stiffness changes during visual stimulation: Differential physiological mechanisms characterized by opposing mechanical effects

Abstract

We performed functional intrinsic Magnetic Resonance Elastography (fiMRE) as well as Time of Flight angiography and BOLD fMRI on 7 healthy human subjects to monitor shear stiffness and arterial dilation during periods of prolonged visual stimulation. FiMRE activation with increased stiffness was observed to occur almost equally within brain white and gray matter ($49.7\pm 14.0\%$ and $40.9\pm 12.2\%$, respectively), while activation with decreased stiffness was significantly ($p=0.018$) more likely to occur in white matter than gray matter ($50.8\pm 11.4\%$ and $37.0\pm 4.5\%$, respectively). At the low mechanical activation and block design frequencies used in this intrinsic MRE (iMRE) approach, the aggregate stiffness change across the entire BOLD activation region was not significant. However, we observed significant reduction in shear stiffness (1.40 ± 0.15 to 0.68 ± 0.22 [kPa], p < 0.001) in areas adjacent to the Posterior Cerebral Artery, where vasodilation is evident, in the V1 region. In addition, we observed significant shear stiffness increase (1.29 ± 0.12 to 0.62 ± 0.16 [kPa], p < 0.001) in areas adjacent to the Middle Temporal or V5 region of the visual cortex. These results show that iMRE can measure intrinsic cerebro-mechanical reactions due to visual stimulation as well as the differential physiological response detected in distinct regions of the visual cortex.

Highlights

• •Different regions of the visual cortex undergo distinct responses to visual stimulation.
• •These responses engage dissimilar reactions from the surrounding cerebrovascular network.
• •This differential cerebrovascular reactivity relies on separate mechanobiological mechanisms that lead to changes in the local cerebral stiffness.
• •Using intrinsic MRE, we characterized these differential physiological mechanisms underlying cerebral stiffness changes.
• •We found significant reductions and increases in cerebral stiffness around the V1 and V5 regions, respectively.

1. Introduction

MR Elastography(MRE) is now well established as a method to measure the mechanical properties in the in vivo brain (V Hiscoxet al, 2016), (Mariappan et al., 2010). By characterizing the elastic response via MR measured deformation fields, MRE provides images of the mechanical properties of soft tissue, such as stiffness, at the resolution of the MR imaging data. Recently, intrinsic MRE (iMRE) methods based on cardiac pressure pulse induced displacement fields have been developed to image viscoelastic properties at frequencies around $1Hz$. (Weaveret al, 2012), (Catheline, 2016) Gordon Wylie et al. (Gordon-Wylieet al, 2018), used quantitative flow (QFLOW) MR imaging in a gelatin phantom with a hydraulic actuation system to imitate the $1Hz$ cardiac cycle and verified the iMRE viscoelastic reconstruction. (Weaveret al, 2012), (Van Houten et al., 2000) One challenge of iMRE in in vivo applications is interpreting the observed viscoelastic signal. While traditional extrinsic MRE (eMRE) depends on the elastic response to peripheral, passively induced displacements at non-physiologic frequencies, iMRE depends on internal, actively generated displacements via the cardiac pressure pulse (i.e. at physiologic frequencies). In this way, there are more possible origins for the viscoelastic properties measured by iMRE than the latent structural response measured in eMRE.

One way to investigate the origins of the elastic response observed in MRE is through the use of a functional imaging approach. This approach has recently been introduced based on extrinsic MRE (eMRE) methods to investigate neuromechanical as a possible source for measured stiffness changes. In a recent study using functional eMRE (feMRE) with a shear wave frequency of $1KHz$, brain biomechanics were quantified using fast time scale imaging to capture neuronal activity (Patzet al, 2019). An overall $10\%$ decrease in stiffness during electrical hind limb stimulation was observed, with stiffness patterns dependent on the cyclic ON/OFF block design interval, which varied from $9s$ to $0.1s$. In another study using simultaneous BOLD fMRI and feMRE, $6-12\%$ stiffness increases within the visual cortex were observed with a shear wave frequency of $60Hz$ and block design visual stimulation intervals of $18s$ to $36s$ (Lan et al., 2020). Multifrequency feMRE experiments by Fehlner et al., at extrinsically induced shear wave frequencies ranging from $25Hz$ to $50Hz$, showed a $2.5\%$ decrease in average whole-brain viscoelastic stiffness. (Fehlner et al., 2014)., (Hirsch et al., 2017)

For iMRE of the in vivo brain, another possible source for the measured viscoelastic signal is cerebrovascular activity. Cerebral vasoreactivity is a mechanism that involves changes in cerebral blood volume (CBV), cerebral blood flow (CBF), cerebral perfusion pressure and cerebral metabolism to maintain blood regulation (Widder et al., 2001). The mechanics behind these factors engage various compartments in the cortex and has been of interest to researchers for some time (K, 2017; Sutera and Skalak, 1993; Fung, 1993). Nevertheless, to date, no imaging method has been developed to measure the differential physiological mechanisms of cerebral vasoreactivity directly. Vasodilation and vasoconstriction, the two main elements of vasoreactivity, are driven by pressure changes in the arterial and capillary bed and smooth muscle cell (SMC) contraction and relaxation within the vessel wall. (Mandeville et al., 2002), (Iadecola, 2017) Parker et al., have translated these effects into a stress relaxation function and described the stress changes during vasodilation and vasoconstriction as two attenuating exponential signals (K, 2017). Likewise, experimentally induced vasodilation in mice via hypercapnic challenge has been shown to decrease brain rigidity as measured by eMRE (Schregel et al., 2017), while overall brain stiffness in humans was shown increase with hypercapnia, as measured by eMRE (Hetzeret al, 2019), (Kreftet al., 2020).

Here, we introduce a novel, non-invasive method for functional iMRE (fiMRE) to characterize the mechanical changes in the brain due to visual stimulation at a low block design frequency. The roughly $1Hz$ displacements originating from the cardiac pressure pulse and the prolonged block design intervals ($300s$) used in this technique enable us to investigate the different quasi-static mechanisms that underlie visual cortex activation.

2. Material and methods

2.1. Image acquisition

This study was performed using seven healthy subjects, 5 male and 2 female, aged 24–32 years old. The following imaging protocol was used for each subject: a T1 anatomic image (voxel size = 0.7986 × 0.7986 × 0.8000 $mm$, FoV = 230 × 230 × 149.6 $mm$); BOLD fMRI using 30 s ON/OFF visual stimulation for 10 cycles (3 mm isotropic voxels, FoV = 240 × 240 × 102 $mm$) to locate the visual cortex; (using the BOLD fMRI as a guide to select a FoV covering the visual cortex) 6 repeated (=3x(ON _5mins+ OFF_5mins) cycles(=30mins total imaging time)) cardiac-gated 4D-QFLOW images with 8 motion phases (Gordon-Wylieet al, 2018) (2.8 $mm$ isotropic voxels, FoV = 200 × 200 × 48 $mm$); 2 (ON/OFF visual stimulation, each ~5mins) Time of Flight (ToF) angiography images (voxel size = 0.625 × 0.625 × 0.650 $m\text{m}$, FoV = 200 × 200 × 144.3 $mm$). Visual stimulation was an 8 Hz full field flickering checker board pattern generated with the $PsychophysicsToolbox$ (http://psychtoolbox.org/) for Matlab (https://www.mathworks.com/). The fiMRE experimental design corresponded to a block design interval of $300$ seconds. All images were acquired on a 3 T Philips Ingenia (https://www.philips.com/global). All participants provided informed written consent and received a complete explanation of the research procedure. The research ethics committee of the CIUSSS de l'Estrie – CHUS, the CÉR (REBA), has examined and approved the consent form and the research protocol (#2016–1341 – DospIRM) in compliance with American standards (FWA #00005894 and IRB #00003849). The data used in this study is available via a formal data sharing agreement with the Université de Sherbrooke.

2.2. Intrinsic MR Elastography reconstruction

The 6 iMRE displacement fields were processed using the subzone based Non Linear Inversion (NLI) algorithm. (Weaveret al, 2012), (Van Houten et al., 2000), (Hiscoxet al., 2018) NLI MRE is based on Navier's equation for time-harmonic motion u, Eq. (V Hiscoxet al, 2016), relating the viscoelastic forces governed by the complex shear modulus, ($\mu ={\mu }_R+i{\mu }_I$) and the $\lambda$ modulus to the inertial forces at the activation frequency, ($f\approx 1Hz$, for cardiac gated motion), governed by the mass density ($\rho =1000\frac{kg}{m^3}$ for water saturated biological tissue). For this study, NLI MRE estimates of the shear modulus distribution were reconstructed using 1.5 $mm$ isotropic voxels and the same imaging volume as the iMRE data (FoV = 200 × 200 × 48 $mm$) (McGarryet al, 2012).

$$\nabla \cdot \mu \left(\nabla \mathbf{u}+\nabla {\mathbf{u}}^T\right)+\nabla \lambda \nabla \cdot \mathbf{u}=-\rho {\omega }^2\mathbf{u}$$ [1]

The processing codes used for this study are protected intellectual property and cannot be shared publicly. The authors are happy to organize the treatment of any suitable data submitted by a qualified researcher for verification purposes.

2.3. Functional Elastography processing

After NLI MRE reconstruction, the six measured storage modulus distributions, ${\mu }_R\left(x\right)=G^{\prime }\left(x\right)$, were separated by visual stimulation state (i.e. ${\mu }_R^{ON}\left(x\right)$ or ${\mu }_R^{OFF}\left(x\right)$) and processed using the Welch's t-test (WELCH, 1947). At each voxel within the imaging volume, the mean and variance were estimated within each group (ON or OFF) and a p-value was calculated for the hypothesis of equal means across the two stimulation states. To identify the activation regions for fiMRE, a post-processing step within each slice was performed to identify regions with 15 or more contiguous voxels each meeting the condition $p-value<0.05$. A Stiffness Factor ($SF$) was calculated for each activation region, $SF=mean({\mu }_R^{ON}\left(x\right))/mean({\mu }_R^{OFF}\left(x\right))$ to indicate the level and nature of shear modulus changes within the activated region. Contrast to noise ratio (CNR) values were also calculated within each activation region, $CNR=(mean({\mu }_R^{ON}\left(x\right))-mean({\mu }_R^{OFF}\left(x\right)))/{\sigma }_{ON,OFF}$. Activation regions with a positive CNR, equivalent to $SF>1$, are identified as positive activation regions, while those with a negative CNR, equivalent to $SF<1$, are identified as negative activation regions. The distribution of these regions between white matter, gray matter and CSF were calculated based on the FAST segmentation tool in the FMRIB Software Library (FSL) (Woolrichet al., 2009). P-values are given for the test of different distributions of positive and negative activation regions between white matter and gray matter. The distribution of activation regions directly within vessel volumes cannot be calculated due to partial volume effects in the 4D-QFLOW data used for iMRE reconstruction.

2.4. ToF processing

ToF images were reconstructed in the same manner as Bernier et al. (2018), to measure arterial diameter with and without visual stimulation. An example of ToF reconstructions and their associated activation region are shown in Fig. 3.

Fig. 3.

Sample fiMRE activation region compared to ToF images for one subject. TOF images are the Posterior Cerebral Artery (PCA) during rest and activation. The red frame shows the $\mathit{P}4$ segment of the PCA, with clearly visible dilation during activation. A negative fiMRE activation region adjacent to the $\mathit{P}4$ is shown in blue. Below, the 3D images highlight the $\mathit{P}4$ dilation during activation (indicated by arrows) and the adjacent activation region. (Note: to emphasize the dilation of the artery and the form of the activation region, these 3D contours are shown from a different angle than the 2D image.).

2.5. Axial and volumetric reprojection

The fiMRE activation regions are relatively small compared to the BOLD fMRI activation maps. In addition, many of the regions of fiMRE activity are associated with changes observed in the neighboring arteries, with the exact geometry of the cerebrovascular tree varying significantly between individuals (Duvernoy et al., 1981). These factors make comparing fiMRE regions across subjects very difficult, as simple registration to standard atlases (such as the MNI atlas (Fonov et al., 2009; Fonovet al., 2011; Collins et al., 1999)) does not ensure that cerebral arteries (or their neighboring regions) will overlay across different subjects. This leads to the co-registered fiMRE activation regions lying close together but not necessarily overlapping in the atlas space. For example, calculating the average number of activations per voxel in atlas space would give a result of roughly $\frac{1}{Nsubj}$ for $Nsubj$ subjects, as no two activation regions actually overlap, despite being close.

To better represent the location of the positive and negative activation regions across multiple subjects, we developed two methods of combining co-registered activation region maps based on projection methods. In the first of these methods, axial reprojection, activation region maps across multiple subjects (co-registered in MNI space) are combined in the axial direction to give the probability of a particular axial slice segment (defined by the number of slices per segment, $Nseg$) containing a region of activation in all of the subjects. The axial reprojection probability for pixel $(i,j)$ in the axial slice plane, ${\Pi }_{ij}^{ax}$, is thus defined as,

$${\Pi }_{ij}^{ax}=\sum _{s=1}^{Nsubj}\sum _{k=1}^{Nslice}A{R}_s(i,j,k)\frac{Nseg}{NsubjNslice},$$ [2]

where $Nsubj$ is the number of subjects, $Nslice$ is the number of slices in the atlas used for co-registration, and $A{R}_s(i,j,k)$ is a binary activation region map for subject $s$ indicating the presence of an activation region at voxel location $(i,j,k)$.

To provide an estimate of the activation region locations across all subjects within the 3D imaging volume, we developed the volumetric reprojection method. The idea behind this method is similar to back projection computed tomography (Michael, 2001), using only the three principal axes of the image coordinates to estimate the envelope shared activation across multiple subjects. The regions identified by the volumetric reprojection method are identified as volumetric activation regions. The details of this method as well as figures showing the axial and volumetric reprojections for the 7 subjects from this study are presented in Appendix 1.

ToF images containing the Posterior and Middle Cerebral Arteries (PCA and MCA, respectively) were analyzed to determine the centerline and diameter of these vessels during rest and visual stimulation (Bernier et al., 2018). After nonlinear transformation into the ToF image space using Advanced Normalization Tools (ANTS) (Avants et al., 2009) and FSL (Woolrichet al., 2009) and subsequent masking of the vessels themselves, fiMRE activation voxels were associated with cerebral artery locations by determining the perpendicular distance from the ToF centerline axis to the nearest fiMRE activation voxel.

3. Results

Fig. 1 shows the $5mm$ Maximum Intensity Projection (MIP) of the volumetric reprojection areas in conjunction with the BOLD fMRI activation maps (see section Axial and volumetric reprojection above). The images are coregistered to the MNI anatomic atlas and each column shows one slice in MNI space (slice number indicated above). The top, middle and bottom rows show the positive volumetric activation regions in green, the negative volumetric activation regions in blue, and the BOLD fMRI regions in red, respectively. As can be seen, portions of the negative volumetric activation regions are within the primary visual cortex (V1) as indicated by the BOLD fMRI region. Conversely, positive volumetric activation regions are outside of the V1 area but reach the MT/V5 area of the visual cortex, marginally overlapping with the BOLD activity.

Fig. 1.

Volumetric reprojection regions and BOLD fMRI activation regions coregistered to the MNI anatomic atlas and shown in slices 63, 67 and 71 of the atlas. The first row shows the 5 mm MIP of the positive volumetric regions (SF > 1) in green. The second row shows the 5 mm MIP of the negative volumetric regions (SF < 1) in blue. The BOLD fMRI regions are shown in red in the third row.

Fig. 2 shows the cerebrovascular atlas (Bernier et al., 2018) and the thresholded volumetric reprojection of the negative ($SF<1$) and positive ($SF>1$) activation regions in blue and green, respectively, from different views. The PCA is only partially visible in the atlas, and certain segments, such as the $P_4$, are absent due to the high inter-subject variability of arterial architecture. Based on the position of the PCA predicted by the cerebrovascular atlas, negative activation regions look to be primarily aligned along the presumptive location of the $P_4$ segment, supporting the idea that negative fiMRE activation is associated with cerebrovascular response to stimulation.

Fig. 2.

A cerebrovascular atlas shown with the volumetric reprojection of fiMRE activity regions from all subjects in axial (a), isometric (b), sagittal (c) and (d) coronal views. The proximal segments of the MCA and PCA can be seen near the center of the atlas. Negative activity regions ($\mathit{SF}<1$) can be seen (in blue) following the $\mathit{P}4$ segment of the PCA, which is not included completely in the atlas. The positive activity regions ($\mathit{SF}>1$) are shown in green.

In general, regions of both increased and decreased ${\mu }_R$ fiMRE activity were observed throughout the brain. Average $SF$ values within the activation regions for the seven participants are shown in Table 1. $SF>1$ and $SF<1$ indicate the positive and negative activation regions, characterized by increased and decreased ${\mu }_R$ values, respectively. These $SF$ measurements imply that intrinsic stiffness increases by a factor of $2.52$ in positive activation regions and decreases by a factor of $0.45$ in negative activation regions during visual stimulation.

Table 1.

Average Stiffness Factors (SF) within the positive and negative activation regions for each subject.

Subject	1	2	3	4	5	6	7	Average
Positive Regions SF	2.43 ± 1.20	1.85 ± 0.63	3.61 ± 1.22	2.49 ± 1.21	2.52 ± 1.43	2.19 ± 1.49	2.54 ± 1.03	2.52 ± 0.54
Negative Regions SF	0.35 ± 0.08	0.50 ± 0.13	0.46 ± 0.17	0.56 ± 0.21	0.64 ± 0.17	0.41 ± 0.13	0.50 ± 0.17	0.48 ± 0.10

In Table 2, stiffness changes are compared within the fiMRE and BOLD activation regions. CNR and p-values are given to indicate the nature and significance of the stiffness changes within the activation regions. The stiffness changes within positive and negative fiMRE regions have significant p-values ($p<0.001$). There is an insignificant, ${\mu }_R$ decrease (CNR = −0.097) within the BOLD fMRI regions (p-value of 0.859). Table 3 shows the percentage distribution of the positive and negative activation regions within the white matter, gray matter and cerebrospinal fluid (CSF). No significant differences were found for the test of different white matter and gray matter distributions between positive and negative activation regions. This table implies that the activation regions are almost equally distributed within white matter and gray matter, with negative regions more likely to be found in the white matter$(p=0.018)$.

Table 2.

Average stiffness changes [kPa] in BOLD fMRI and fiMRE activation regions.

	BOLD fMRI	Positive Regions	Negative Regions
mean$\left({\mu }_R^{ON}\right)$ [kPa]	0.56 ± 0.16	1.29 ± 0.12	0.68 ± 0.22
mean$\left({\mu }_R^{OFF}\right)$ [kPa]	0.57 ± 0.16	0.62 ± 0.16	1.40 ± 0.15
p-value	0.85	p < 0.001	p < 0.001
CNR	−0.10	4.39	−3.62

Table 3.

Percentage distribution [%] of positive and negative fiMRE activation regions within different brain components.

	White Matter	Gray Matter	p-value
Positive Regions [%]	49.7 ± 14.0	40.9 ± 12.2	0.23
Negative Regions [%]	50.8 ± 11.4	37.0 ± 4.5	0.02

Overall, the fiMRE activation regions were located close to cerebral arteries, based on the distance to arterial centerlines determined as per Bernier et al. (2018),. Over 60% of activated voxels were within $2-3mm$ of cerebral arteries of diameter $0.75mm$ or less. We also observed regions that did not appear to be related to any artery, with two typical examples of this shown in Appendix 1.

As noted in Bizeau et al. (Bizeauet al., 2018), the PCA undergoes noticeable vasodilation during visual stimulation. Fig. 3 shows co-registered ToF (OFF (left) and ON (right)) and fiMRE images for one subject within the region of the PCA. A negative activation region with decreased ${\mu }_R$ is shown in blue. The red frame in Fig. 3 illustrates the apex of the $P_4$ segment of the PCA, which becomes apparent in the ON image on the right, indicating its dilation during visual activation. The average ${\mu }_R^{OFF}$ and ${\mu }_R^{ON}$ values that led to the identification of the region of interest shown in Fig. 3, as well as an example motion field from one instance of the MRE motion imaging sequences during visual activation are shown in Fig. 4.

Fig. 4.

Images (a) and (b) show the mean ${\mathit{\mu }}_{\mathit{R}}^{\mathit{OFF}}$ and mean ${\mathit{\mu }}_{\mathit{R}}^{\mathit{ON}}$ values, respectively, within a slice plane in the neighborhood of the activation region shown in Fig. 3. The exact outline of the fMRE activation region is highlighted in blue. Image (c) shows the measured motion vectors from one instance of the repeated MRE measurements during visual stimulation.

The average dilatations across seven subjects observed within the negative and positive volumetric reprojection regions (shown in Fig. 1) as a percentage of total average vessel diameter were 1.21% and 0.35%, respectively. This measurement is possible using the combined reprojection regions, which do not exclude vessel regions for individual subjects. Dilation values are reported as a percentage of the total average vessel diameter within each subject, as per Eq. (Weaveret al, 2012).

4. Discussion

In this study we present a non-invasive iMRE method to measure the elastic changes in the active brain at frequencies related to the natural cardiac cycle (Weaveret al, 2012). This novel technique is able to identify two different cerebrovascular activations within the visual cortex that lead to regions of increased and decreased cerebral stiffness. Furthermore, we are able to characterize the mechanobiology of these different physiological mechanisms by differential relationships between cerebral stiffness changes and vasodilation. Unlike other methods, the low frequency nature of this iMRE based approach allows a special sensitivity to the cerebrovascular response to visual stimulation, which is significantly different than the BOLD fMRI response.

Characterizing the mechanical properties of the brain provides essential information for comprehending brain function and its pathologies. FiMRE is novel in the use of intrinsic activation for the MRE displacement field, but similar studies based on feMRE have been performed. Fehlner et al. (2014), (Hirsch et al., 2017) performed multifrequency feMRE in humans with visual stimulation and observed a $2.5\%$ decrease in overall brain stiffness. In a more recent study in mice, Patz et al. (Patzet al, 2019), performed feMRE ($1KHz$ actuation) with electrical stimulation of the hind limb at three different time scales: slow, fast and ultra-fast corresponding to $9s$, $0.9s$ and $100ms$ stimulus intervals, respectively. While activation regions differed at each time-scale, stiffness decreases were observed in the OFF state of the experiment compared to the ON and control states. At these fast imaging time scales, the suggested mechanism for stiffness changes are neuromechanical responses during stimulation. In a recent study using extrinsic activation MRE, Lan et al. (2020), performed simultaneous BOLD fMRI and feMRE ($60Hz$ actuation) with visual stimulation on humans and observed localized increases in stiffness ($6$ to $11\%$) within the visual cortex. The fiMRE method presented here differs significantly from these previous studies in the use of very low frequencies for both visual stimulation ($300s$ block design) and mechanical activation (cardiac cycle, $\approx 1Hz$), which correspond to different mechanobiological mechanisms, such as vasoreactivity, than the neuromechanical responses observed during higher frequency studies.

We observed activation regions with reduced and increased stiffness due to visual stimulation. These regions were almost equally distributed in the white and gray matter (Table (Weaveret al, 2012)), although negative activation regions were significantly ($p=0.02$) more likely to occur in white matter than gray matter. Thomas et al. (2014) have found that the nature of cerebrovascular reactivity differs between GM and WM. Some type of varied response is reasonable, and the mechanisms that underlie the positive and negative stiffness responses observed here may also vary. While significant stiffness changes were observed within the fiMRE activation regions, and some of these regions overlap with BOLD activation, the aggregate stiffness change across the entire BOLD activation region was not significant (Table 2). The absence of fiMRE activation regions within BOLD-activated regions may be explained by a lack of vasoreactive response in the veins, where bold signal is observed.

In a recent study (Solamen et al., 2019), phantoms fabricated with two different gelatin concentrations were tested with iMRE methods ($\omega \approx 1Hz$) using poroelastic and viscoelastic models. The Full-Width Half-Maximum of the Edge Spread Function (FWHM-ESF) of the measured stiffness shows that the transition from the border of the stiffer gelatin to the softer gelatin occurs over a $12mm$ step. Due to the long wavelengths associated with iMRE (McGarry et al., 2019), very fine structures such as arteries($\approx 0.1-10mm$ in diameter) may not be localized exactly, but their effective stiffness changes should be visible within a distance roughly corresponding to the FWHM-ESF observed by Solamen et al. (2019), which is well within the distance range indicated by the activated voxel distributions observed here.

Vasoreactivity is a complex mechanism that involves mechanical changes in the vessels and the surrounding tissue. Primary dilation of the arteries has been shown to be followed by a drop in pressure (Mandeville et al., 2002) and is one potential mechanism for observed stiffness decreases. The viscoelastic response, described by the stress relaxation function developed by Parker et al. (K, 2017), implies a $40\%$ shear stress decrease ($2.5KPa$ to $1KPa$, equivalent to $SF=0.4$) in less than $100$ seconds after vasodilation. This is in good agreement with our finding that the average $SF$ for negative regions ($SF<1$) is $0.45$. This relaxation continues for $700$ seconds until it reaches a plateau at a shear stress of roughly $0.5KPa$.

Another related potential mechanism for the observed stiffness reduction during visual cortex activation is the shear stress based myogenic response of smooth muscle cells (SMC) that permits vasodilation to take place (Iadecola, 2017). The relaxation of these SMCs is known to change the viscoelastic properties of the vessel wall (Spronck et al., 2012). This dilation has been shown to propagate upstream from the capillary bed to the supplying arteries (Longdenet al, 2017). Katharina Schregel et al. (2017), performed extrinsic MRE ($\omega =1KHz$) while inducing vasodilation by hypercapnic challenge in healthy mice and observed a significant decrease in cortical rigidity due to vasodilation.

Fig. 1 indicates that positive fiMRE activation regions occurred around the area of the Middle Temporal ($MT$) or $V5$ region of the visual cortex, associated with motion recognition. (Watsonet al, 1993), (Wandell et al., 2007) Sadaghiani et al. (2009) performed BOLD fMRI and ASL with static and flickering checkerboard visual stimulations and showed $MT$/$V5$ activation with the flickering stimulation (as used in this study) while no such activation was observed with the static stimulation. In addition, they demonstrated a similarity between BOLD fMRI and ASL signals in both experiments, where ASL represents the CBF changes. CBF increases in the $MT$/$V5$ area have been observed during stimulation (Kim et al., 1999), while Huber et al. (Huberet al., 2014), showed nearly $0\%$ change in CBV in the $MT$/$V5$ area. Mechanically, this increase in blood flow while blood volume remains constant would indicate higher local pressure gradients, a potential mechanism for observed stiffness increases. For this study, the average dilatations within the negative and positive volumetric reprojection regions shows that the amount of blood volume change due to arterial dilation was nearly $4X$ smaller within the positive fiMRE volumetric reprojection regions than in the negative regions, although small amounts of dilation (0.35% on average) were observed within this region. In addition, Fig. 1 shows small areas of BOLD activation within the proximity of the $MT$/$V5$ region, indicating increases in CBF as observed in the studies cited above.

The asymmetry of the fMRE response in the region of the PCA is apparent in Fig. 1. Fig. 2 shows that the asymmetry observed in these negative regions (SF < 1) is not a total asymmetry but only a diminished effect in one branch of the PCA. This could indicate differing roles of the two PCA branches in supplying oxygenated blood to the V1 region during stimulation (Roje-Bedeković et al., 2012), (Matthias et al., 1996).

Other mechanisms that could explain stiffness increases during visual stimulation have also been observed, as discussed above (Lan et al., 2020). The relatively low stiffness changes observed attributed to neurological effects, (Patzet al, 2019), (Lan et al., 2020) as well as the vastly different experimental block designs associated with these responses, would limit the impact of these effects on the iMRE results used here.

One limitation of this study was the lack of a cerebrovascular model incorporated into our viscoelastic reconstruction. Such a model could be used to distinguish cerebrovascular flow effects from effects truly related to tissue mechanics, such as SMC activation or neuron mechanobiology. There are also limitations due to the use of the iMRE technique, which relies on a low Velocity ENCoding sensitivity (VENC) to measure the tissue motion component throughout the imaging volume (Lotz et al., 2002). For voxels containing partial volumes of fluid (blood) and tissue, low VENC allows potentially higher vascular flow velocities to distort the overall voxel velocity measurement, and this to varying degrees due to phase-wrap effects. In general, these effects were likely minimized by the use of a t-test statistic to determine the activation regions. Thus, only voxels that showed significant stiffness changes with respect to their variation across $6$ repetitions were considered. Due to local variations in cerebrovascular flow velocities and phase wrapping, these mixed volumes would likely have shown high variation across repeats. In addition, there are other intrinsic motion sources such as respiration, CSF effects, subject motion, etc. which were not considered in this study. Another limitation is due to the long imaging time for the iMRE sequence that limited the number repeats available for functional imaging. For the future studies, the iMRE sequence could be accelerated to enable additional ON and OFF repetitions with the same image quality.

The volumetric reprojection method used in this study merits discussion. This method was developed to address the challenge of combining cerebrovascular anatomy across multiple subjects, particularly within the region of the $P_4$ segment of the PCA. This effect is partly responsible for lack of definition of this region within the cerebrovascular atlas shown in Fig. 2 and presented in (Bernier et al., 2018). One existing method to address this challenge is the dilation approach, illustrated in a simple 2D example in Fig. 5, where neighboring regions are dilated until they overlap (Fig. 5.1), and the overlapping region is then identified as the combined region of interest (Fig. 5.2). As can be seen, a disadvantage of this approach is that the identified combined region tends not to coincide with the initial regions. Fig. 6 shows the same 2D regions treated by the volumetric reprojection approach, where the initial regions (Fig. 6.1) are projected along the axes (Fig. 6.2 and 6.3), and then reprojected into the 2D volume (Fig. 6.4) and then thresholded to identify the combined region of interest (Fig. 6.5). As can be seen, using this approach, the identified combined region tends to coincide with the original regions. A fully developed 3D example of this approach using actual activation regions from this study is presented in Appendix 1.

Fig. 5.

A 2D example of the dilation method for identifying combined regions of interest. Individual regions (shown in Fig. 6. (1)) are first dilated (1) and then the overlapping area is used and the combined region of interest (2).

Fig. 6.

A 2D example of the volumetric reprojection method for identifying combined regions of interest. Individual regions (1) are projected along the axes (2 & 3) and then reprojected into the volume (4). The resulting image is thresholded to identify the combined region of interest (5).

5. Conclusions

Different regions of the visual cortex undergo distinct responses to visual stimulation. These responses engage dissimilar reactions from the surrounding cerebrovascular network. This differential cerebrovascular reactivity relies on separate mechanobiological mechanisms that lead to changes in the local cerebral stiffness. By using the intrinsic cerebral motions caused by the cardiac pressure pulse as well as advanced MR Elastography imaging methods, we characterized these differential physiological mechanisms underlying cerebral stiffness changes during visual stimulation. We found significant reductions in cerebral stiffness around the V1 region, where vasodilation was evident, as well as significant increases in cerebral stiffness around the V5 region. We also observed inverse relationships between stiffness changes and vasodilation within these two regions. These stiffness changes consistent with previously observed cerebrovascular effects.

Functional imaging based on low-frequency iMRE (fiMRE) allows us to monitor brain activation responses that are different than those observed with BOLD fMRI. FiMRE is a potential tool to better comprehend the stiffness changes in the active brain. Here we have demonstrated a direct quantitative measurement of the stiffness changes associated with differential mechanisms for vasoreactivity in biologically plausible areas. This new technique could eventually open up new horizons for understanding cerebral vasoreactivity and its associated brain disorders.

Author statement

Reihaneh Forouhandehpour: Conceptualization, Software, Validation, Formal analysis, Investigation, Data curation, Writing – original draft, Visualization, Project administration, Methodology. Michaël Bernier: Software, Writing – review & editing, Methodology. Guillaume Gilbert: Conceptualization, Writing – review & editing, Methodology. Russell Butler: Conceptualization, Software, Writing – review & editing, Methodology. Kevin Whittingstall: Conceptualization, Methodology, Writing – review & editing, Supervision. Elijah Van Houten: Conceptualization, Methodology, Software, Formal analysis, Data curation, Writing – original draft, Writing – review & editing, Supervision, Funding acquisition

Funding

The authors gratefully acknowledge the essential support of: The Natural Sciences and Engineering Research Council of Canada (NSERC); Compute Canada; Calcul Québec; Réseau de Bio-Imagerie du Québec (RBIQ) and the Centre de recherche du CHUS (CRCHUS). These funding sources had no involvement in the study design; in the collection, analysis and interpretation of data; in the writing of the report; and in the decision to submit the article for publication.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary data

The following is the Supplementary data to this article: