From structure to mechanics: exploring the role of axons and interconnections in anisotropic behavior of brain white matter

According to various experimental studies, the role of axons in the brain's white matter (WM) is still a subject of debate: Is the role of axons in brain white matter (WM) limited to their functional significance, or do they also play a pivotal mechanical role in defining its anisotropic behavior? Micromechanics and computational models provide valuable tools for scientists to comprehend the underlying mechanisms of tissue behavior, taking into account the contribution of microstructures. In this review, we delve into the consideration of strain level, strain rates, and injury threshold to determine when WM should be regarded as anisotropic, as well as when the assumption of isotropy can be deemed acceptable. Additionally, we emphasize the potential mechanical significance of interconnections between glial cells-axons and glial cells-vessels. Moreover, we elucidate the directionality of WM stiffness under various loading conditions and define the possible roles of microstructural components in each scenario. Ultimately, this review aims to shed light on the significant mechanical contributions of axons in conjunction with glial cells, paving the way for the development of future multiscale models capable of predicting injuries and facilitating the discovery of applicable treatments.

On the application of hybrid deep 3D convolutional neural network algorithms for predicting the micromechanics of brain white matter

BACKGROUND

Material characterization of brain white matter (BWM) is difficult due to the anisotropy inherent to the three-dimensional microstructure and the various interactions between heterogeneous brain-tissue (axon, myelin, and glia). Developing full scale finite element models that accurately represent the relationship between the micro and macroscale BWM is however extremely challenging and computationally expensive. The anisotropic properties of the microstructure of BWM computed by building unit cells under frequency domain viscoelasticity comprises of 36 individual constants each, for the loss and storage moduli. Furthermore, the architecture of each unit cell is arbitrary in an infinite dataset.

METHODS

In this study, we extend our previous work on developing representative volume elements (RVE) of the microstructure of the BWM in the frequency domain to develop 3D deep learning algorithms that can predict the anisotropic composite properties. The deep 3D convolutional neural network (CNN) algorithms utilizes a voxelization method to obtain geometry information from 3D RVEs. The architecture information encoded in the voxelized location is employed as input data while cross-referencing the RVEs' material properties (output data). We further develop methods by incorporating parallel pathways, Residual Neural Networks and inception modulus that improve the efficiency of deep learning algorithms.

RESULTS

This paper presents different CNN algorithms in predicting the anisotropic composite properties of BWM. A quantitative analysis of the individual algorithms is presented with the view of identifying optimal strategies to interpret the combined measurements of brain MRE and DTI.

SIGNIFICANCE

The proposed Multiscale 3D ResNet (M3DR) algorithm demonstrates high learning ability and performance over baseline CNN algorithms in predicting BWM tissue properties. The hybrid M3DR framework also overcomes the significant limitations encountered in modeling brain tissue using finite elements alone including those such as high computational cost, mesh and simulation failure. The proposed framework also provides an efficient and streamlined platform for implementing complex boundary conditions, modeling intrinsic material properties and imparting interfacial architecture information.

Systematic analysis of constitutive models of brain tissue materials based on compression tests

It's crucial to understand the biomechanical properties of brain tissue to comprehend the potential mechanisms of traumatic brain injury. This study, distinct from homogeneous models, integrates axonal coupling in both axial and transverse compressive experiments within a continuum mechanics framework to capture its intricate mechanical behaviors. Fresh porcine brains underwent unconfined compression at strain rates of 0.001/s and 0.1/s to 0.3 strain, allowing for a comprehensive statistical analysis of the directional, regional, and strain-rate-dependent mechanical properties of brain tissue. The established constitutive model, fitted to experimental data, delineates material parameters providing intuitive insights into the stiffness of gray/white matter isotropic matrices and neural fibers. Additionally, it predicts the mechanical performance of white matter matrix and axonal fibers under compressive loading. Results reveal that gray matter is insensitive to loading direction, exhibiting insignificant stiffness variations within regions. White matter, however, displays no significant differences in mechanical properties under axial and transverse loading, with an overall higher average stress than gray matter and a more pronounced strain-rate effect. Stress-strain curves indicate that, under axial compression, white matter axons primarily resist the load before transitioning to a matrix-dominated response. Under transverse loading, axonal fibers exhibit weaker resistance to lateral pressure. The mechanical behavior of brain tissue is highly dependent on loading rate, region, direction, and peak strain. This study, by combining experimentation with phenomenological modeling, elucidates certain phenomena, contributing valuable insights for the development of precise computational models.

An Ogden hyperelastic 3D micromechanical model to depict Poynting effect in brain white matter

Shear and torsional load on soft solids such as brain white matter purportedly exhibits the Poynting Effect. It is a typical nonlinear phenomenon associated with soft materials whereby they tend to elongate (positive Poynting effect) or contract (negative Poynting effect) in a direction perpendicular to the shearing or twisting plane. In this research, a novel 3D micromechanical Finite Element Model (FEM) has been formulated to describe the Poynting effect in bi-phasic modeled brain white matter (BWM) representative volume element (RVE) with axons tracts embedded in surrounding extracellular matrix (ECM) for simulating brain matter's response to pure and simple shear. In the presented BWM 3D FEM, nonlinear Ogden hyper-elastic material model is deployed to interpret axons and ECM material phases. The modeled bi-phasic RVEs have axons tied to the surrounding ECM. In this proof-of-concept (POC) FEM, three simple shear loading configurations and a pure shear case were analyzed. Root mean square deviation (RMSD) was calculated for stress and deformation response plots to understand the effect of axon-ECM orientations and loading conditions on the degree of Poynting behavior. Variations in normal stresses (S11, S22, or S33) perpendicular to the shear plane underscored the significance of axonal fiber-matrix interactions. From the simulated ensemble of cases, a transitional dominance trend was noticed, as simple sheared axons showed pronounced Poynting behavior, but shear deformation build-up in the purely sheared brain model exhibited the highest Poynting behavior at higher strain % limits. At lower strain limits, simple shear imparted across and perpendicular to axonal tract directions emerged as the dominant Poynting effect configurations. At high strains, the stress-strain% plots manifested mild strain stiffening effects and bending stresses in purely sheared axons, substantiated the strong non-linearity in brain tissues' response.

Harmonic viscoelastic response of 3D histology-informed white matter model

White matter (WM) consists of bundles of long axons embedded in a glial matrix, which lead to anisotropic mechanical properties of brain tissue, and this complicates direct numerical simulations of WM viscoelastic response. The detailed axonal geometry contains scales that range from axonal diameter (microscale) to many diameters (mesoscale) imposing an additional challenge to numerical simulations. Here we describe the development of a 3D homogenization model for the central nervous system (CNS) that accounts for the anisotropy introduced by the axon/neuroglia composite, the axonal trace curvature, and the tissue dynamic response in the frequency domain. Homogenized models that allow the incorporation of all the above factors are important for accurately simulating the tissue's mechanical behavior, and this in turn is essential in interpreting non-invasive elastography measurements. Geometric and material parameters affect the material properties and thus the response of the brain tissue. More complex, orthotropic, or anisotropic material properties are to be considered as necessitated by the 3D tissue structure. An assembly of micro-scale 3D representative elemental volumes (REVs) is constructed, leading to an integrated mesoscale WM finite element model. Assemblies of microscopic REVs, with orientations based on geometrical reconstructions driven by confocal microscopy data are employed to form the elements of the WM model. Each REV carries local material properties based on a finite element model of biphasic (axon-glial matrix) unidirectional composite. The viscoelastic response of the microscopic REVs is extracted based on geometric information and fiber volume fractions calculated from the relative distance between the local elements and global axonal trace. The response of the WM tissue model is homogenized by averaging the shear moduli over the total volume (thus deriving effective properties) under realistic external loading conditions. Under harmonic shear loading, it is proven that that the effective transverse shear moduli are higher than the axial moduli when the axon moduli are higher than the glial. Methodologically, the process of using micro-scale 3D REVs to describe more complex axon geometries avoids the partition process in traditional composite finite element methods (based on partition of finite element grids) and constitutes a robust algorithm to automatically build a WM model based on available axonal trace information. Analytically, the model provides unmatched simulation flexibility and computational power as the position, orientation, and the magnitude of each tissue building block is calculated using real tissue data, as are the training and testing processes at each level of the multiscale WM tissue.

Sensitivity analysis of effective transverse shear viscoelastic and diffusional properties of myelinated white matter

Motivated by the need to interpret the results from a combined use of in vivo brain Magnetic Resonance Elastography (MRE) and Diffusion Tensor Imaging (DTI), we developed a computational framework to study the sensitivity of single-frequency MRE and DTI metrics to white matter microstructure and cell-level mechanical and diffusional properties. White matter was modeled as a triphasic unidirectional composite, consisting of parallel cylindrical inclusions (axons) surrounded by sheaths (myelin), and embedded in a matrix (glial cells plus extracellular matrix). Only 2D mechanics and diffusion in the transverse plane (perpendicular to the axon direction) was considered, and homogenized (effective) properties were derived for a periodic domain containing a single axon. The numerical solutions of the MRE problem were performed with ABAQUS and by employing a sophisticated boundary-conforming grid generation scheme. Based on the linear viscoelastic response to harmonic shear excitation and steady-state diffusion in the transverse plane, a systematic sensitivity analysis of MRE metrics (effective transverse shear storage and loss moduli) and DTI metric (effective radial diffusivity) was performed for a wide range of microstructural and intrinsic (phase-based) physical properties. The microstructural properties considered were fiber volume fraction, and the myelin sheath/axon diameter ratio. The MRE and DTI metrics are very sensitive to the fiber volume fraction, and the intrinsic viscoelastic moduli of the glial phase. The MRE metrics are nonlinear functions of the fiber volume fraction, but the effective diffusion coefficient varies linearly with it. Finally, the transverse metrics of both MRE and DTI are insensitive to the axon diameter in steady state. Our results are consistent with the limited anisotropic MRE and co-registered DTI measurements, mainly in the corpus callosum, available in the literature. We conclude that isotropic MRE and DTI constitutive models are good approximations for myelinated white matter in the transverse plane. The unidirectional composite model presented here is used for the first time to model harmonic shear stress under MRE-relevant frequency on the cell level. This model can be extended to 3D in order to inform the solution of the inverse problem in MRE, establish the biological basis of MRE metrics, and integrate MRE/DTI with other modalities towards increasing the specificity of neuroimaging.

On the Transversely Isotropic, Hyperelastic Response of Central Nervous System White Matter Using a Hybrid Approach

A numerical and experimental hybrid approach is developed to study the constitutive behavior of the central nervous system white matter. A published transversely isotropic hyperelastic strain energy function is reviewed and used to determine stress–strain relationships for three idealized, simple loading scenarios. The proposed constitutive model is simplified to a three-parameter hyperelastic model by assuming the white matter's incompressibility. Due to a lack of experimental data in all three loading scenarios, a finite element model that accounts for microstructural axons and their kinematics is developed to simulate behaviors in simple shear loading scenarios to supplement existing uniaxial tensile test data. The parameters of the transversely isotropic hyperelastic material model are determined regressively using the hybrid data. The results highlight that a hybrid numerical virtual test coupled with experimental data, can determine the transversely isotropic hyperelastic model. It is noted that the model is not limited to small strains and can be applied to large deformations.

Nonlinear mechanics of soft composites: hyperelastic characterization of white matter tissue components

This paper presents a bi-directional closed-form analytical solution, in the framework of nonlinear soft composites mechanics, for top-down hyperelastic characterization of brain white matter tissue components, based on the directional homogenized responses of the tissue in the axial and transverse directions. The white matter is considered as a transversely isotropic neo-Hookean composite made of unidirectional distribution of axonal fibers within the extracellular matrix. First, two homogenization formulations are derived for the homogenized axial and transverse shear moduli of the tissue, based on definition of the strain energy density function. Next, the rule of mixtures and Hashin-Shtrikman theories are used to derive two coupled nonlinear equations which correlates the tissue shear moduli to these of its components. Closed-form solutions for shear moduli of the components are then obtained by solving these equations simultaneously. In order to validate the hyperelastic characteristics of components obtained in previous step, they are used in a bottom-up approach in a micromechanical model of the tissue with the aim of predicting the directional homogenized responses of the tissue. Comparison of model predictions with the experimental test results reported for corona radiata and corpus callosum white matter structures reveals very good agreements with the experimental results in both directions. The model predictions are also in good agreement with the analytical solution obtained by the iterated homogenization technique. Results indicate that axonal fibers are almost ten times stiffer than the extracellular matrix under large deformations.

Axons Embedded in a Tissue May Withstand Larger Deformations Than Isolated Axons Before Mechanoporation Occurs

Diffuse axonal injury (DAI) is the pathological consequence of traumatic brain injury (TBI) that most of all requires a multiscale approach in order to be, first, understood and then possibly prevented. While in fact the mechanical insult usually happens at the head (or macro) level, the consequences affect structures at the cellular (or microlevel). The quest for axonal injury tolerances has so far been addressed both with experimental and computational approaches. On one hand, the experimental approach presents challenges connected to both temporal and spatial resolution in the identification of a clear axonal injury trigger after the application of a mechanical load. On the other hand, computational approaches usually consider axons as homogeneous entities and therefore are unable to make inferences about their viability, which is thought to depend on subcellular damages. Here, we propose a computational multiscale approach to investigate the onset of axonal injury in two typical experimental scenarios. We simulated single-cell and tissue stretch injury using a composite finite element axonal model in isolation and embedded in a matrix, respectively. Inferences on axonal damage are based on the comparison between axolemma strains and previously established mechanoporation thresholds. Our results show that, axons embedded in a tissue could withstand higher deformations than isolated axons before mechanoporation occurred and this is exacerbated by the increase in strain rate from 1/s to 10/s.

A three-dimensional micromechanical model of brain white matter with histology-informed probabilistic distribution of axonal fibers

This paper presents a three-dimensional micromechanical model of brain white matter tissue as a transversely isotropic soft composite described by the generalized Ogden hyperelastic model. The embedded element technique, with corrected stiffness redundancy in large deformations, was used for the embedment of a histology-informed probabilistic distribution of the axonal fibers in the extracellular matrix. The model was linked to a multi-objective, multi-parametric optimization algorithm, using the response surface methodology, for characterization of material properties of the axonal fibers and extracellular matrix in an inverse finite element analysis. The optimum hyperelastic characteristics of the tissue constituents, obtained based on the axonal and transverse direction test results of the corona radiata tissue samples, indicated that the axonal fibers were almost thirteen times stiffer than the extracellular matrix under large deformations. Simulation of the same tissue under a different loading condition, as well as that of another white matter tissue, i.e., the corpus callosum, in the axonal and transverse directions, using the optimized hyperelastic characteristics revealed tissue responses very close to those of the experiments. The results of the model at the sub-tissue level indicated that the stress concentrations were considerably large around the small axons, which might contribute into the brain injury.

Micromechanics of brain white matter tissue: A fiber-reinforced hyperelastic model using embedded element technique

A transverse-plane hyperelastic micromechanical model of brain white matter tissue was developed using the embedded element technique (EET). The model consisted of a histology-informed probabilistic distribution of axonal fibers embedded within an extracellular matrix, both described using the generalized Ogden hyperelastic material model. A correcting method, based on the strain energy density function, was formulated to resolve the stiffness redundancy problem of the EET in large deformation regime. The model was then used to predict the homogenized tissue behavior and the associated localized responses of the axonal fibers under quasi-static, transverse, large deformations. Results indicated that with a sufficiently large representative volume element (RVE) and fine mesh, the statistically randomized microstructure implemented in the RVE exhibits directional independency in transverse plane, and the model predictions for the overall and local tissue responses, characterized by the normalized strain energy density and Cauchy and von Mises stresses, are independent from the modeling parameters. Comparison of the responses of the probabilistic model with that of a simple uniform RVE revealed that only the first one is capable of representing the localized behavior of the tissue constituents. The validity test of the model predictions for the corona radiata against experimental data from the literature indicated a very close agreement. In comparison with the conventional direct meshing method, the model provided almost the same results after correcting the stiffness redundancy, however, with much less computational cost and facilitated geometrical modeling, meshing, and boundary conditions imposing. It was concluded that the EET can be used effectively for detailed probabilistic micromechanical modeling of the white matter in order to provide more accurate predictions for the axonal responses, which are of great importance when simulating the brain trauma or tumor growth.

Characterization of the three-dimensional kinematic behavior of axons in central nervous system white matter

Traumatic injury to axons in white matter of the brain and spinal cord occurs primarily via tensile stretch. During injury, the stress and strain experienced at the tissue level is transferred to the microscopic axons. How this transfer occurs, and the primary constituents dictating this transfer must be better understood to develop more accurate multi-scale models of injury. Previous studies have characterized axon tortuosity and kinematic behavior in 2-dimensions (2-D), where axons have been modeled to exhibit non-affine (discrete), affine (composite-like), or switching behavior. In this study, we characterize axon tortuosity and model axon kinematic behavior in 3-dimensions (3-D). Embryonic chick spinal cords at different development stages were excised and stretched. Cords were then fixed, transversely sectioned, stained, and imaged. 3-D axon tortuosity was measured from confocal images using a custom-built MATLAB script. 2-D kinematic models previously described in Bain et al. (J Biomech Eng 125(6):798, 2003) were extended, re-derived, and validated for the 3-D case. Results showed that 3-D tortuosity decreased with stretch, exhibiting similar trends with changes in development as observed in the 2-D studies. Kinematic parameters also displayed similar general trends. Axons demonstrated more affine behavior with increasing stretch and development. In comparison with 2-D results, a smaller percentage of the populations of 3-D axons were predicted to follow pure non-affine behavior. The data and kinematic models presented herein can be incorporated into multi-scale CNS injury models, which can advance the accuracy of the models and improve the potential to identify axonal injury thresholds.

Cellular biomechanics of central nervous system injury

With a growing interest in how the brain responds and remodels itself following a traumatic injury, this chapter outlines the major organizing principles of how to study these injuries in the laboratory and extend these findings back into the clinic. A new repertoire of models is available to examine the response of isolated circuits of the brain in vitro, and to study precisely how mechanical forces applied to even small regions of these circuits can disrupt the entire circuit dysfunction. We review the existing knowledge garnered from these models and our current understanding of mechanically sensitive receptors and channels activated immediately following trauma. In turn, we point to the emergence of in silico models of network function that will lead to an improved understanding of the principles for the remodeling of circuit structure after traumatic, possibly pointing out new biological rules for circuit reassembly that would help guide new therapies for reconstructing brain circuits after trauma.

The mechanics of traumatic brain injury: a review of what we know and what we need to know for reducing its societal burden

Traumatic brain injury (TBI) is a significant public health problem, on pace to become the third leading cause of death worldwide by 2020. Moreover, emerging evidence linking repeated mild traumatic brain injury to long-term neurodegenerative disorders points out that TBI can be both an acute disorder and a chronic disease. We are at an important transition point in our understanding of TBI, as past work has generated significant advances in better protecting us against some forms of moderate and severe TBI. However, we still lack a clear understanding of how to study milder forms of injury, such as concussion, or new forms of TBI that can occur from primary blast loading. In this review, we highlight the major advances made in understanding the biomechanical basis of TBI. We point out opportunities to generate significant new advances in our understanding of TBI biomechanics, especially as it appears across the molecular, cellular, and whole organ scale.

Finite Element Modeling of CNS White Matter Kinematics: Use of a 3D RVE to Determine Material Properties

Axonal injury represents a critical target area for the prevention and treatment of traumatic brain and spinal cord injuries. Finite element (FE) models of the head and/or brain are often used to predict brain injury caused by external mechanical loadings, such as explosive waves and direct impact. The accuracy of these numerical models depends on correctly determining the material properties and on the precise depiction of the tissues' microstructure (microscopic level). Moreover, since the axonal microstructure for specific regions of the brain white matter is locally oriented, the stress, and strain fields are highly anisotropic and axon orientation dependent. Additionally, mechanical strain has been identified as the proximal cause of axonal injury, which further demonstrates the importance of this multi-scale relationship. In this study, our previously developed FE and kinematic axonal models are coupled and applied to a pseudo 3-dimensional representative volume element of central nervous system white matter to investigate the multi-scale mechanical behavior. An inverse FE procedure was developed to identify material parameters of spinal cord white matter by combining the results of uniaxial testing with FE modeling. A satisfactory balance between simulation and experiment was achieved via optimization by minimizing the squared error between the simulated and experimental force-stretch curve. The combination of experimental testing and FE analysis provides a useful analysis tool for soft biological tissues in general, and specifically enables evaluations of the axonal response to tissue-level loading and subsequent predictions of axonal damage.

A micromechanical procedure for viscoelastic characterization of the axons and ECM of the brainstem

In this study, the optimal viscoelastic material parameters of axon and extracellular matrix (ECM) in porcine brain white matter were identified using a genetic algorithm (GA) optimization procedure. The procedure was combined with micromechanical finite element analysis (FEA) of brain tissue and experimental stress relaxation tests on brainstem specimens to find the optimal material coefficients of axon and ECM. The stress relaxation tests were performed in tension on 10 brainstem specimens at 3% strain level. The axonal volume fraction in brainstem was measured from the Scanning Electron Microscopy images of the brain tissue. A square periodic volume element was selected to represent the microscale homogenized brainstem tissue. Periodic boundary conditions were applied on the square volume element to mimics the repetitive nature of the volume element. Linear viscoelastic material properties were assumed for the brain tissue constituents under small deformation. The constitutive behavior was expressed in terms of Prony series. The GA procedure searched for the optimal material parameters by fitting the time-dependent tissue stresses of brain tissue FEA to the stresses of relaxation tests under the same loading conditions. The optimization procedure converged after 60 iterations. The initial elastic modulus of axon was found to be 12.86kPa, three times larger than that of ECM. The long-term elastic modulus of axon was 3.7kPa, while for ECM this value was 1.03kPa. The concordance correlation coefficient between FEA estimated elastic modulus of brainstem tissue using the optimal material properties and the experimental elastic modulus of brainstem specimens was 0.952, showing a strong agreement. The optimal material properties of brain tissue constituents can find applications in micromechanical analysis of brain tissue to gain insight into diffuse axonal injures (DAIs).