Assessment of vibration modulated regional cerebral blood flow with MRI

Human brain experiences vibration of certain magnitude and frequency during various physical activities such as vehicle transportation and machine operation, which may cause traumatic brain injury or other brain diseases. However, the mechanisms of brain pathogenesis due to vibration are not fully elucidated due to the lack of techniques to study brain functions while applying vibration to the brain at a specific magnitude and frequency. Here, this study reported a custom-built head-worn electromagnetic actuator that applied vibration to the brain in vivo at an accurate frequency inside a magnetic resonance imaging scanner while cerebral blood flow (CBF) was acquired. Using this technique, CBF values from 45 healthy volunteers were quantitatively measured immediately following vibration at 20, 30, 40 Hz, respectively. Results showed increasingly reduced CBF with increasing frequency at multiple regions of the brain, while the size of the regions expanded. Importantly, the vibration-induced CBF reduction regions largely fell inside the brain's default mode network (DMN), with about 58 or 46 % overlap at 30 or 40 Hz, respectively. These findings demonstrate that vibration as a mechanical stimulus can change strain conditions, which may induce CBF reduction in the brain with regional differences in a frequency-dependent manner. Furthermore, the overlap between vibration-induced CBF reduction regions and DMN suggested a potential relationship between external mechanical stimuli and cognitive functions.


Introduction
Vibration is a common mechanical stimulus to our brain during daily life and work, such as in-vehicle transportation and machine operation, which has been strongly associated with symptoms of motion sickness ( Donohew and Griffin, 2004 ;Howarth and Griffin, 2003 ;Joseph and Griffin, 2007 ;Murdin et al., 2011 ;Weech et al., 2018 ), vibration syndrome ( Chang et al., 1994 ), vertigo, and dizziness ( Zwergal and Dieterich, 2020 ). In more severe cases, both short-period dynamic high impact and long-period repetitive moderate vibration can cause trau-Abbreviations: DMN, Default Mode Network; CBF, Cerebral Blood Flow; TBI, Traumatic Brain Injury; pCASL, pseudo-Continuous Arterial Spin-Labeling; MRE, Magnetic Resonance Elastography; ROI, Region Of Interest. Curry et al., 2002Curry et al., , 2005Hua et al., 2017 ;Palmer and Collin, 1993 ;Stoyneva et al., 2003 ). And the vasoconstriction has also been found to be dependent on both the magnitude and frequency of the vibration ( Bovenzi et al., 2000 ;Ye and Griffin, 2014 ). On the other hand, patients with TBI also show reduced cerebral blood flow (CBF), a major physiological marker of brain functions, in the bilateral frontotemporal regions, together with structural changes of the cerebral blood vessels and disintegrity of white matter ( Clark et al., 2017 ;Raichle, 2015 ;Wang et al., 2015 ). Widespread deficits of CBF have also been found in gray matter in correlation with severity of TBI and cognitive dysfunction ( Ware et al., 2020 ).
Recent computational simulation studies appeared to indicate that vibration-induced effects on brain functions may well be frequencydependent. Specifically, a study used finite element method (FEM) to computationally simulate the motion and strain for brain models based on dynamics data from injured and non-injured athletes subjected to head impacts and found that the brain was more susceptible to vibration at frequency ranging between 20 and 40 Hz, with strain peaking at around 30 Hz ( Laksari et al., 2018 ). However, all the present imaging modalities in neuroscience prevent in vivo quantitative measurement of physiological and functional changes in the brain during vibration because the head motion is always considered an artifact and supposed to be carefully avoided during imaging or eliminated in post-hoc imaging analysis. Therefore, it remains a fundamental challenge to experimentally characterize the in vivo frequency-dependence of vibrationinduced effects on brain functions despite its critical importance for understanding the mechanism of brain disease caused by vibration.
To overcome this challenge and prove our hypothesis that exposure to the vibration on the brain could reduce CBF in the brain in a regionspecific and frequency-dependent fashion, a custom-built electromagnetic actuator was worn by each subject inside the MR head coil to introduce mechanical vibration to the brain at frequencies of 20, 30, and 40 Hz, respectively, and CBF was measured immediately after vibration. Using this technique, we found that vibration caused reduction of CBF, which was highly dependent on the frequency of vibration as well as the region in the brain. In particular, the regions with vibrationreduced CBF appeared to largely overlap with the brain's default mode network (DMN) associated with cognitive processes ( Yeshurun et al., 2021 ). These findings not only prove the concept of characterizing brain functions under specific vibration conditions, but also provide an important relationship between the external mechanical stimulation and the internal CBF and cognitive functions of the human brain, which may be essential for fully understanding the pathophysiological mechanisms of brain damage due to exposure to environmental vibrations as well as potentially help developing better approaches for preventing/treating vibration-induced brain injury.

Subjects and ethical approval
Fifty-two healthy volunteers (age 25.87 ± 2.78 years old, 31F/21 M) participated in this study. All subjects provided informed consent, as approved by the Science and Technology Ethics Committees of the Shanghai Jiao Tong University. Among all volunteers, 7 subjects' data were excluded due to motion artifacts, resulting in a total of 45 subjects (26.05 ± 3.09 years old, 26F/19 M) for subsequent analyses.

Study design
To investigate the CBF changes after vibration, we carried out imaging in two sessions ( Fig. 1 ): a reference session without vibration (session 1) and with vibration (session 2). Session 1 included a high-resolution T1-weighted (T1W) imaging sequence for anatomical MRI (aMRI), followed by a 3D pseudo-continuous arterial spin-labeling (pCASL) sequence to measure resting CBF (C 0 ). At the end of session 1, subjects would rest for 5 min. Session 2 included 3 similar imaging modules corresponding to imaging at 3 vibration frequencies (20 Hz, 30 Hz, and 40 Hz). The excitation power of the actuator was set to 3.3 W in all frequency. The order of the measurement frequency for each module was carried out in a randomized, counterbalanced way. The vibration motion was induced using a high-frequency accuracy electromagnetic actuator ( Qiu et al., 2021 ). The vibration amplitude of the actuator was measured by an accelerometer at 20 Hz, 30 Hz, and 40 Hz, with a magnitude of 1.97 ± 0.15 mm, 1.62 ± 0.11 mm, and 1.42 ± 0.10 mm, respectively. Extensive tests have been carried out in previous studies to verify that there is no interference between the actuation and the imaging ( Qiu et al., 2021 ). The vibration actuator stayed on the heads of the subjects for the whole experiment. It was not taken off during the rest modules. The subjects stayed inside the scanner until all the scans were finished.
At each imaging module of session 2, wave images from vibration were also acquired using a magnetic resonance elastography (MRE) sequence to characterize the in-brain displacement field of vibration at 3 frequencies. After vibration at each imaging module of session 2, CBF was measured immediately using a pCASL sequence. To ensure the reliability and repeatability of the results, the above two imaging steps were repeated for each module. The total time for both sessions was 90 min. Subjects were requested to keep their eyes open during the experiment.
Additional 10 subjects were recruited to verify post vibration if the CBF returns to baseline after 5-minute rest. The baseline regression includes 3 frequency blocks ( Fig. 1 c). In each block, resting-state CBF (R1, R3, R5) values were first measured 5 min before vibration. After vibration, the subjects had a 5-minute rest. Then, CBF (R2, R4, R6) values were measured again. Paired student t-tests were used to verify whether the CBF went back to the baseline between different vibrational modules (R1 vs. R2; R3 vs. R4; R5 vs. R6).
MRE was performed during each vibration session ( Qiu et al., 2021 ). A single-shot echo-planar imaging (EPI) based MRE sequence with a first-nulling motion encoding gradient (MEG) was used for brain displacement measurement. The MEG had a strength of 30 mT/m and the repetition/echo time (TR/TE) was 3400/71.6 ms. A total of 34 slices with a thickness of 4 mm was collected for each object. The FOV was 224 mm × 224 mm with a matrix size of 128 × 128. The principal component of the displacement fields was filtered by a 2D Gaussian filter with a kernel size of 5 × 5 pixels and a standard deviation of 0.65. The three principal strain components ( 1 , 2 , and 3 ) and octahedral shear strain ( ) for each frequency were calculated ( McGarry et al., 2011 ) to explore their correlation with the changes of CBF.

Fig. 1.
Experimental set-up and procedure for measuring the vibration effect on cerebral blood flow. (a) Custom-built head worn electromagnetic vibration actuator consisting several plates that transmitted vibration to the brain. The plates were attached with soft sponge pads to ensure the scanning was comfortable. The actuator was used for both vibration actuation and MRE acquisition ( Qiu et al., 2021 ). A supine position was adopted for all imaging procedures. (b) The imaging protocol includes two sessions. Anatomical images (T1W) and CBF maps in the absence of vibration were acquired as references in session 1. In session 2, CBF was measured immediately after the vibration using a pCASL sequence. During the vibration modules, MRE sequence were utilized to characterize the in-brain displacement field. The CBF measurements were repeated once more to ensure experimental reproducibility. The order of frequency scanning between 20, 30, 40 Hz was randomized to minimize the potential influence of sequential measurement order. (c) Baseline regression tests were designed to verify if the CBF returned to baseline after a 5-minute vibration at 20, 30, and 40 Hz. In each block, CBF maps were acquired before vibration and after the 5-minute rest.

Data analyses
First, skulls were excluded from T1W images using BET2 ( Smith, 2002 ) from the FMRIB Software Library ( Jenkinson et al., 2012 ;Smith et al., 2004 ;Woolrich et al., 2009 ) (FSL v6.0, Oxford center for Functional MRI of the Brain, Oxford, UK). Then, CBF maps were realigned to T1W images using SPM12 (Wellcome Trust center for Neuroimaging, London, UK,). CBF maps were resampled into 2 mm 3 isotropic voxels and normalized to the anatomical standard space defined by the Montreal Neurological Institute (MNI) using Advanced Normalization Tools (ANTs) ( Avants et al., 2011 ). Masks from the automated anatomical labeling atlas (AAL) ( Tzourio-Mazoyer et al., 2002 ) were also used. Cortical surface ROIs defined by AAL template in MNI space are given in the Supplement Information Table S1. CBF maps were standardized for reducing individual variance by = ( − ) ( Liu et al., 2016a ;Pfefferbaum et al., 2011 ), where is the standardized CBF value at each voxel, is the original CBF value, and are the mean and standard deviation of the whole-brain CBF. Finally, images were smoothed with a Gaussian kernel of 8 × 8 × 8 mm 3 full width at half maximum (FWHM) by SPM12.
Significance analyses were carried out concerning the normalized CBF maps using paired student t-tests or two-sample t-tests. A two-tailed permutation test with 5000 permutations was used to find significant clusters with the tool Permutation Analysis of Linear Models (PALM) ( Winkler et al., 2016 ), and multiple comparisons were corrected by the family-wise error rate (FWER) method with threshold-free cluster en- hancement (TFCE) ( Smith and Nichols, 2009 ). A cluster-based threshold of Z > 2.3 was applied. The significance level was set to p < 0.05 for the suprathreshold clusters. The cluster analyses were thresholded at t > 1.96 or t < − 1.96. Toolbox for Data Processing & Analysis of Brain Imaging (DPABI_V6.0 toolkit ,) and MATLAB (MathWorks, Natick, MA, United States) was used for all statistical analysis. Then, the number of overlapped significant voxels in each brain network (MNI space) was estimated for each frequency. The percentage of overlapped significant clusters in all networks and all significant regions were also calculated respectively. In addition, paired student t-tests were made to compare the mean CBF values at the resting state and each of the vibration state at each network for all subjects.

Vibration decreased CBF with increasing frequency
During session 1, brain anatomy (T1W images) and CBF (3D-ASL) in the absence of vibration were acquired as references. During session 2, after vibration (3D-ASL) at frequency of 20, 30, 40 Hz, CBF were immediately acquired twice. These frequencies were chosen because of their significant role in inducing TBI ( Laksari et al., 2018 ). Fig. 2 a-b shows typical anatomical T1W images and of CBF maps measured at either the resting state (RS) or immediately after vibration at 20, 30, 40 Hz, respectively. For each frequency, the measurement of CBF were repeated twice to ensure experimental reproducibility. No significant differences were observed in any clusters at the same frequency (paired student t-tests, p > 0.05). Therefore, the mean values of the CBF maps at the same frequency from the two repeated measurements were used for analysis. Fig. 2 c shows magnified view of CBF maps in the corresponding boxed region in the top row CBF maps of Fig. 2 a-b, of which the intensity appeared to be decreasing with increasing frequency. Statistical analysis results confirmed that CBF in the boxed region (posterior cingulate cortex, PCC; precuneus, PCUN) decreased significantly in the presence of vibration with increasing frequency. ( Fig. 2 d, paired student t-tests, RS vs. 20 Hz vs. 30 Hz vs. 40 Hz, p < 0.05). For the baseline re-gression session, no significant differences were observed in any clusters at each of the frequency block (R1 vs. R2, R3 vs. R4, R5 vs. R6) (paired student t-tests, p > 0.05). These results indicate that CBF could go back to the baseline between different vibrational modules after a 5-minute rest. Fig. 3 displays in comparison the regional distribution of CBF variations either in the absence of vibration (resting state) or in the presence of vibration at frequency of 20, 30, 40 Hz, respectively (paired student ttest). Regions of interest (ROIs) showing clusters with significant group differences ( p < 0.05) were illustrated.
Comparing CBF maps acquired in the presence of vibration at various frequencies, we found significant differences of CBF in ROIs between vibrations at 20 Hz and 40 Hz. Specifically, the significant decrease of CBF occurred in the region of ANG, SMG, SFG, TPOmid, TPOsup, IFG, MFG, and inferior frontal gyrus, orbital (ORBinf) (p corr < 0.05, Fig. 3 c). However, no significant difference of CBF was observed between the experimental group with vibration at either 30 Hz versus 40 Hz, or 20 Hz versus 30 Hz. Fig. 3 d shows that the cluster size of the CBF reduction regions increased from 0 to about 1.5 × 10 4 as the vibration frequency increased from 20 to 40 Hz. Detailed statistics of the cluster reports are given in the Supplement Information Table S2. Furthermore, typical displacement fields and at 3 frequencies were shown in Supplementary Figures S1 and S2. The mean and standard deviation of the displacement due to vibration were 13.56 ± 2.87 m (20 Hz), 13.01 ± 2.97 m (30 Hz), and 12.76 ± 4.03 m (40 Hz), respectively. Paired t-tests were used to compare the mean values of 1 , 2 , 3 , and at regions with significant CBF changes ( Figs. 3 a, b) across subjects for 30 Hz and 40 Hz. Significant differences were observed for 1 ( p < 0.05), 2 ( p < 0.01) and ( p < 0.0001).

Vibration-induced CBF reduction regions overlapped with the brain's default mode network
To explore how CBF changes could affect brain function, we compared the brain regions showing significant CBF reduction with the 7 brain networks known for cognitive functional connectivity ( Yeo et al., 2011 ). The number of overlapping voxels between the regions with significant CBF reduction and those in each of the 7 brain network was estimated for each frequency. We found a large overlap between the regions with significant CBF reduction and the Default Mode Network (DMN) in the brain after vibration at either 30 Hz ( Fig. 4 a) or 40 Hz ( Fig. 4 b). Quantitatively, 58% of voxels in the significant region were in DMN, accounting for 9.03% of DMN in the brain after vibration at 30 Hz ( Fig. 4 c), and 45.64% of voxels in the significant CBF reduction region were in DMN, accounting for 20.02% of DMN in the brain after vibration at 40 Hz ( Fig. 4 d). Detailed statistics of the overlapped significant CBF reduction regions are shown in Supplement Information Table  S3.
In addition, we calculated the mean CBF at each network for all subjects after vibration to the brain at each frequency. Paired student t-tests showed no significant differences of CBF in each network between each vibration frequency (Supplement Information Figure S3), but CBF in the subregions (PCC.L and ACG.R) in DMN was significantly different after vibration at different frequency (Supplement Information Figure S4), possibly due to the average effects of CBF in DMN.

Discussions
Physical stimuli of our head in daily life are remarkably diverse, posing great challenges to the understanding brain functions across physical conditions. A vibration is a common form of head motion. Elucidating the effects of vibration on brain functions not only expands our understanding of the basic science of the brain, but also provides new evidence for the mechanism of vibration-induced brain injury. In this study, we investigated the effects of brain vibration on CBF at three different frequencies that are known to be associated with brain injury. We found that compared to the resting state, dynamic vibration that produces oscillating displacement and changes of strain distributions (Supplement Fig. 4. Overlap of the regions with significant reduction of CBF after vibration with the seven brain functional connectivity networks. (a-b) Voxels in regions with significant CBF reduction (SR) after vibration at 30 Hz, 40 Hz, respectively, were overlaid with the seven functional connectivity networks of the brain. (c-d) Pie charts of the percentage of voxels in the significant CBF reduction region that overlapped with the seven brain networks after vibration at 30 Hz, 40 Hz, respectively. Information Figure S1 and Figure S2) could induce distinct regional reduction of CBF, which was frequency-dependent, i.e., the regions with CBF reductions showed progressively lower CBF at increasing frequency of the vibration. This finding is consistent with that vibration causes an early effect of vasoconstriction in HAVS, which is frequency-dependent ( Bovenzi et al., 2000 ;Ye and Griffin, 2014 ). We also found that the regions with significant CBF reduction after vibration fell inside DMN that is associated with the brain's cognitive functional connectivity.
Previous studies have shown that CBF can be reduced significantly after brain injury in cases such as mild TBI and sports-related concussion ( Churchill et al., 2017 ;Ge et al., 2009 ;Lin et al., 2016 ;Meier et al., 2015 ;Peng et al., 2016 ;Wang et al., 2019 ). Our finding of significant CBF reduction after vibration was consistent with those clinical observations. The method reported in the present study suggests that the vibrationbased quantitative measurement technique has the potential to simulate impact and vibration to the human brain while accessing the physiological responses of the brain. Therefore, this method may be very useful for investigation of various vibration-induced brain diseases.
Furthermore, we found in this study that the regions with significant CBF reduction were mainly located in PCC and PCUN within DMN. These regions are known to be associated with attention, memory, perception, action, and cognition ( Gordon et al., 2020 ;Liang et al., 2013 ;Raichle, 2015 ;Warren et al., 2014 ;Yu et al., 2019 ). Such close relationship between the brain regions that responded to vibration by changing CBF and those associated with cognitive functions provides tangible clues to explain why people feel tired after staying in a working condition with vibrations such as riding in car or operating a percussion drill. These results are also in line with other studies concerning CBF and brain functions of TBI patients. For example, a study showed evidence for increased cerebrovascular reactivity and function connectivity in the medial regions of the DMN in college athletes experienced concussion ( Militana et al., 2016 ). Another study reported that the first 5-mins psychomotor vigilance task decreased the CBF of mild TBI patients in the acute phase in DMN areas ( Liu et al., 2016b ).
To understand the effect of vibration on cognitive resources, N-back tests ( Gevins and Cutillo, 1993 ;Owen et al., 2005 ) were carried out to evaluate the changes of cognitive resource based on working memory. The test protocol was explained in detail in supplements. Results showed that vibration as a physical stimulation could induce alteration of the cognitive resources. Studies have already showed neural activity was related with CBF changes ( Buxton, 2021 ;Claassen et al., 2021 ;Jones et al., 2004 ). As one of the limitations of this study, the role of cognitive resources was not investigated here, but merits a future study. Therefore, our results bring attention to the fact that vibration as a physical stimulation may induce cognition, memory, or action-related changes.
Our finding of the frequency-dependence of CBF reduction due to vibration has many implications. In the experiments with vibration at frequency of 20, 30, 40 Hz, the size of the regions with significant CBF reduction consecutively expanded when the frequency exceeded 20 Hz. Although we did not test it, we speculated a consistent trend of expanding CBF reduction regions as the frequency further increased. This indicated that in an environment with vibration, absorption or damping for vibration at frequency over 20 Hz is desired to protect the brain. Compared with computational studies that showed peak strain in the brain exposed to vibration at 30 Hz, our results indicated that during an impact, the brain might respond by varying CBF in different patterns. Moreover, the fact that CBF reduction in specific regions such as PCC and PCUN in a frequency-dependent fashion may open a new door for using physical modulation of CBF with external vibration as a therapeutic approach to treat certain brain regions. Another indication of our results is that the vibration may influence the measurements of CBF or other perfusion-related values. Therefore, in terms of measurement order, it is suggested that CBF or BOLD-fMRI tests be carried out before applying any physical stimulation to the brain. We observed that CBF changes with strain, and the strain values of 40 Hz were significantly higher than that of 30 Hz, with showing the most significant differences (Supplementary Figure S5). Future work includes further clarification of the role of strain in affecting CBF changes, and investigation of the interrelationship between frequency, strain, and CBF in patients.
The limitations of this study were: (1) Only three frequencies were investigated due to the allowable scan time; (2) Considering the limit of scan time, a single delay pCASL sequence was used with PLD = 1800 ms ( Alsop et al., 2015 ), which may be improved with muti-PLD pCASL to improve the quantification of CBF; (3) Considering the intrinsically low SNR of ASL, to avoid the potential caveat of statistical effect, future work should consider other reference techniques such as positron emission computed tomography (PET) to measure the CBF changes after vibration to verify the effect reported. (4) Only N-back tests were used to assess the cognitive changes after vibration in a preliminary way, no detailed behavioral and cognitive changes were investigated. In future study, the protocol for human model studies should be further improved and also expanded to include patients with mild TBI or depression. It should also be studied the delicate application of vibration with respect to specific regions for accurate control of CBF modulation.

Conclusion
It was discovered in this study that vibration to the human brain that brings tissue strain changes caused increasing reduction of CBF with increasing vibration frequency, and the regions in which vibration modulated CBF largely overlapped with those responsible for functional connectivity. Such relationship between the vibration-induced CBF modulation and the brain functional structure provides a new way to understand the functional influences of mechanical stimuli to the human brain. And the method presented in this report can also be used to apply mechanical vibration to the human brain as a stimulus to modulate CBF in studies of various vibration-related brain diseases.

Ethics statement
The studies involving human participants were reviewed and approved by the Science and Technology Ethics Committees of the Shanghai Jiao Tong University. The patients/participants provided their written informed consent to participate in this study.

Declaration of Competing Interests
The author(s) declared no potential conflicts of interest concerning the research, authorship, and/or publication of this article.

Data availability
The original contributions presented in the study are included in the article/supplementary material, Further inquiries can be directed to the corresponding author(s).