Computational brain models are a useful tool to assess the dynamic strain response of the brain during head impact. Population-average brain models are typically used, but their predictions depend on many factors (e.g. geometry, material properties) that differ between the model and an individual. However, subject-specific (SS) models can be developed with model inputs and validation targets acquired through specialized magnetic resonance imaging (MRI) sequences that allow for the acquisition of anatomical structure, computation of material properties using magnetic resonance elastography (MRE), and measurement of in-vivo brain deformation using tagged MRI (tMRI). The goal of this study is to create and evaluate individualized computational brain models using SS anatomy, material properties, and experimental brain deformation. The objectives are 1) to assess if linear viscoelastic (LVE) material properties from MRE can be directly implemented in brain models to improve validation performance without further calibration, and 2) investigate the use of MRE to assign material properties based on regions of similar mechanical properties rather than similar anatomy.Experimental brain data from a 31yo male was acquired using conventional structural images (T1w, T2w, and diffusion), MRE at three frequencies (30, 50, 70 Hz), and tMRI under axial rotation of the head (3 rad/s, 200 rad/s2). All data were co-registered and interpolated to the same isotropic resolution (1.5 mm). A SS 3D brain model was created from the segmented anatomical images, including 7 brain regions, ventricles, falx/tentorium, sub-arachnoid space, and skull. Storage and loss moduli from MRE scans were used to estimate the LVE properties of the brain regions using a global optimization scheme that utilized the average of MRE moduli within a brain region, as well as published dynamic shear material testing data outside of the MRE frequency range (0.1 – 150 Hz). The literature testing data was normalized to the magnitude of the MRE data to extrapolate the SS data without affecting the estimated stiffness. Material parameters were estimated for two types of heterogeneity: A) separating the brain into 7 regions by segmented labels (anatomical heterogeneity), and B) separating the brain into 10 regions by stratifying the complex modulus estimated from MRE (mechanical heterogeneity).Three models, all with SS anatomy, were simulated in this study: 1) a ‘general’ material model using properties from a previously validated model, 2) a SS material model with anatomical heterogeneity (‘SS-anat’), and 3) a SS material model with mechanical heterogeneity (‘SS-mech’). All models were simulated for 100 ms using the material point method with the tagged MRI loading conditions. The tMRI experiments were used to compute 3D displacement and strain fields at a temporal resolution of 18 ms. Evaluation of the models to the tMRI data was conducted using global strain measures (maximum principal strain (MPS), maximum axonal strain (MAS), and volume fraction) as well as voxel-level comparison using the Correlation and Analysis Score (CORA).The global strain shows an improvement of MPS (90th, 95th, 99th) between the ‘general’ (0.092, 0.112, 0.142) and ‘SS-anat’ (0.042, 0.048, 0.061) models in comparison to the experimental data (0.037, 0.043 0.05). The same trend is observed in global MAS. The magnitude of volume fraction of the brain exceeding an MPS threshold (2%, 4%, 6% MPS) or exceeding an MAS threshold (1%, 2%, 3% MAS) shows a better agreement for the ‘SS-anat’ model. The voxel-level CORA evaluations (calculated for each voxel and averaged for all brain regions) shows a similar trend with slightly better scores for the ‘SS-anat’ model (0.50 for E_xy, 0.70 for MPS, and 0.43 for fiber strain). The ‘SS-anat’ model and ‘SS-mech’ model showed similar scores for the global and local comparisons.The inclusion of SS LVE properties from MRE improved the model validation performance, and further calibration using independent tMRI data or material testing data may provide further improvement. There was little difference between the anatomical and mechanical heterogeneity models, but this needs additional study and perhaps improved calibration procedures.

Subject-Specific 3d Brain Simulations Using Heterogenous, Linear Viscoelastic Material Properties Derived From Magnetic Resonance Elastography

Paper Proceedings: https://asmedigitalcollection.asme.org/IMECE/IMECE2021/volume/85543

ASME 2021 International Mechanical Engineering Congress and Exposition (IMECE2021)

Supporting the need among researchers worldwide for additional online and multimedia content, ASME now offers the ASME Conference Video Collection. This platform will collect and provide long-term access to conference presentation videos, encompassing all technology areas within ASME's technical conference and events program. About 30,000-40,000 engineers across all mechanical engineering and related disciplines participate in ASME conferences each year. This includes authors of approximately 6,000 to 8,000 technical papers published by ASME, the majority of whom also submit video presentations for our virtual conferences. The video collection will be available as an add-on to existing product packages or as a separate subscription.