Personalized Finite Element Modeling of the Brain: Understanding Subject-Specific Deformation Patterns

Author: ORCID icon orcid.org/0000-0003-1074-4890
Giudice, Sebastian, Mechanical and Aerospace Engineering - School of Engineering and Applied Science, University of Virginia
Advisor:
Panzer, Matthew, EN-Mech/Aero Engr Dept, University of Virginia
Abstract:

Traumatic brain injuries (TBI) are a significant source of injury, disability, and death. Central to the investigation of TBI is the development and use of finite element (FE) models of the brain, which provide an in-depth assessment of the brain’s mechanical response during impact to the head. Historically, these models have been developed to represent the anatomy of a specific target demographic, usually the 50th percentile male. This presents a significant limitation as structural differences between two brains under the same impact conditions may lead to different mechanical responses. However, the relationships between the neuroanatomy and biomechanical response of the brain are poorly understood. Furthermore, current FE brain model results cannot be compared to subject-specific neuroimaging diagnostics used to evaluate the effects of TBI on the structure and function of the brain. Therefore, the goals of this dissertation were to develop a novel methodology for automatically generating anatomically detailed subject-specific brain models, and to use these models to investigate how variation in the neuroanatomy influences the biomechanical response of the brain.

To facilitate large-scale biomechanical investigations of the relationships between the neuroanatomy and brain deformation, a registration-based morphing (RBM) framework was developed, which leveraged image registration operations required to geometrically align a template brain image to the anatomy of a given subject. By applying these linear and nonlinear mathematical operations to a corresponding template brain FE model, morphed subject-specific models were generated. Since nonlinear transformations were utilized, the subject-specific models accurately represented the size, shape, and local anatomy of the brain. This technique was computationally efficient, automatic, and robust.

The foundation of the RBM framework developed in this dissertation was the template image and corresponding brain model. A new symmetric template brain image, called the CAB-20MSym template, was constructed from magnetic resonance imaging (MRI) scans of 20 healthy adult males. A corresponding template brain model was developed to exactly represent its anatomy. Material properties of the CAB-20MSym template model were based on heterogeneous stiffness maps from magnetic resonance elastography (MRE) data, and nonlinear viscoelastic properties were calibrated using subject-specific models of specimens previously used for an in-situ brain deformation study. Using an independent set for brain deformation data, the CAB-20MSym template model was shown to demonstrate a biofidelic response.

To investigate the relationships between the neuroanatomy and brain deformation, 100 subject-specific models were developed using RBM, and simulated under loading conditions that represented a concussive American football head impact. In parallel, the anatomical characteristics of these subjects were quantified using mesh-based and neuroimaging measures of the anatomy. These included intracranial volume, brain dimensions, regional volumes, and principal components explaining the variance in the size, shape, and local anatomy of the brain. Finally, a set of statistical models relating global brain strain and anatomical characteristics were trained using various linear regression techniques, including simple linear regression, multiple linear regression, principal component regression, and partial least squares regression. Across all regression models, intracranial volume was found to be the strongest predictor of maximum principal strain (MPS-95), demonstrating a strong linear relationship. Despite this clear relationship, there was overwhelming evidence that suggested that the local anatomy of the skull and brain influenced the deformation response of the brain. Collectively, these statistical models represent the most comprehensive assessment of the link between anatomy and brain deformation and serve both as tools for predicting MPS-95 and to provide insight on how an individual’s neuroanatomy relates to global brain deformation.

The work presented in this dissertation provides a substantial contribution to the TBI biomechanics community, facilitating large-scale investigations on how the neuroanatomy influences brain biomechanics and enabling research across diverse populations that have been typically neglected by the TBI biomechanics community. On a broader scale, RBM bridges the gap between neuroimaging and biomechanics research, enabling direct comparisons between biomechanical results and neuroimaging diagnostics for TBI in humans. Ultimately, this research enables the interdisciplinary investigations required to further the fundamental understanding of TBI mechanisms and its effects on the brain. This knowledge will be crucial for improving injury diagnosis, treatment, and prevention.

Degree:
PHD (Doctor of Philosophy)
Keywords:
Traumatic Brain Injury, Injury Biomechanics, Personalized Medicine, Brain Deformation, Image Registration
Language:
English
Rights:
All rights reserved (no additional license for public reuse)
Issued Date:
2020/12/02