Introduction

Viscoelastic properties of the human brain tissue vary slightly during hypercapnia-induced vasodilation1. The primary fMRI contrast mechanisms rely on neuronal-induced vasodilation response2. However, the mechanical effects of vasomotion in cortical structures are neither investigated nor considered in high resolution fMRI studies yet1,3. Even minor tissue displacements on the fMRI sub-voxel level can lead to erroneous relative fMRI signal changes of the same magnitude, indicating the importance of investigating these mechanical effects during vasomotion on the resulting fMRI signal which to our knowledge has not been studied systematically to date. Using biomechanical simulations, we investigated the influence of variations in brain tissue mechanical property, due to vasodilation on high resolution fMRI images (published recently4). First, we performed finite element simulation of brain tissue deformations in a gyrus of the human brain using ultra-high resolution histology data as input. Second, we simulated high-resolution fMRI datasets at multiple spatial resolutions and contrasts in order to investigate the effect of the estimated tissue deformations on the resulting fMRI signal.

Methods

Anatomical Model
We constructed a two-dimensional geometric model of a coronal slice of the right middle frontal gyrus obtained from ultra-high resolution histology data of 20 µm from the HumanBrainProject5. The six cortical layers in GM were manually traced. The input geometry consisting of surrounding GM, WM, and CSF was imported in COMSOL Multiphysics (V.5.5,COMSOL Multiphysics Göttingen) for finite element simulations. The biomechanics of CSF were not included in the simulations.

Biomechanical Simulations
The brain tissue in our study was modeled as neo-Hookean hyperelastic and assumed to be isotropic6 and nearly incompressible7. Literature values8,9 of brain tissue properties were used as input to the simulation model. Biomechanical simulations were performed to study the effect sizes of local volume increase, stiffness increase and the combination of both. In the first simulation scenario, the shear modulus of a segment within the first cortical layer as highlighted in red in Fig. 1 underwent an increase of up to 10% of its initial value in 20 equidistant steps of 0.5%. In the second scenario, the same segment underwent an increase in volume by uniformly extending its lateral dimensions in 0.5% steps. Finally, the increases in stiffness and volume were combined in a single simulation to study their combined effect.

Functional MRI Simulations
The gyrus geometry obtained from the biomechanical simulations was imported to MRiLab (http://mrilab.sourceforge.net/) to estimate the sensitivity of different high-resolution fMRI sequences to tissue displacement caused by vasodilation. The simulations included the three most commonly used fMRI variants: proton density contrast as used in arterial spin labeling (ASL), Vascular-Space-Occupancy (VASO) and blood-oxygen-level-dependent (BOLD) on a 7T MRI system and idealized RF coils and B0 homogeneity. The fMRI simulation parameters were taken from the literature10 and mapped to the mechanical gyrus model.
The magnitude MR images were constructed from the simulated k-space data at three 0.25, 0.5 and 1.0 mm isotropic resolutions for the simulated gyral geometries for volume increase.
From the corresponding MR images of the modeled gyrus in the resting and "activated" states, voxel-wise maps of the relative signal change caused by tissue displacement were calculated.

Results

Resulting tissue displacement for three simulation scenarios i.e. increase in stiffness, volume and the combined effects are shown in Fig. 2. One can see that the volume increase simulation resulted in a larger displacement compared to the stiffness change. The largest displacement was observed for the combined effect. For visualization purposes a scaling factor of 5 was chosen for the figure. Fig. 3 shows the cortical gyrus geometry in the resting state as used for fMRI simulations for idealized example VASO, BOLD and ASL protocols in different resolutions. The tissue displacement due to a 2% volume increase resulted in spurious fMRI signal changes. The results yielded the strongest signal change in methods based on images with high contrast between neighboring tissues and cortical layers. It should also be noted that the artifacts in signal changes increase with decreasing voxel size.

Discussion and Conclusion

Our biomechanical simulation results showed remarkable tissue displacement resulting from blood flow-induced alterations in mechanical tissue properties. In our simulated fMRI measurements, these displacements resulted in significant artifacts. Our findings highlight the importance of considering and characterizing tissue displacement as a potential source of fMRI artifacts. Although our simulation study requires validation, it provides insights into potential biomechanical effects and how they may impact high resolution fMRI measurements.