MR Imaging of Human Brain Mechanics In Vivo: New Measurements to Facilitate the Development of Computational Models of Brain Injury

Philip V. Bayly, Ahmed Alshareef, Andrew K. Knutsen, Kshitiz Upadhyay, Ruth J. Okamoto, Aaron Carass, John A. Butman, Dzung L. Pham, Jerry L. Prince, K.t. Ramesh, Curtis L. JohnsonAnnals 2021 · DOI

TL;DR

This paper introduces a novel approach using in vivo MRI to measure the biomechanical response of the living human brain to skull accelerations. This dataset, publicly available, aims to bridge the gap between animal/cadaveric models and human brain mechanics, providing crucial data for developing, calibrating, and validating computational models of traumatic brain injury (TBI).

What Problem Does This Paper Address?

Traumatic brain injury (TBI) is a significant health concern, and computational models are vital for understanding its mechanisms and developing protective measures. However, these models have historically relied on data from animal models or cadavers, which differ significantly from the living human brain. This discrepancy limits the accuracy and predictive power of TBI models. Furthermore, directly measuring brain deformation in living humans during injurious impacts is ethically impossible. This paper addresses this gap by proposing and demonstrating a method to non-invasively measure the biomechanical response of the living human brain under sub-injury levels of skull acceleration, providing a much-needed dataset for TBI model development.

Methods

The study employs advanced Magnetic Resonance Imaging (MRI) techniques to capture various aspects of brain mechanics:

  • Tagged MRI: This technique is used to measure the brain's response to impulsive skull loading. By modulating magnetization within the tissue, temporary patterns are created that move with the brain, allowing for the tracking of 3D displacements and the calculation of strain fields over time. This method provides full-field measurements of dynamic brain deformation.
  • Magnetic Resonance Elastography (MRE): MRE is utilized to measure the brain's response to harmonic skull motion across a range of frequencies. This technique uses an external driver to create shear waves in the soft tissue, and the speed of these waves is used to estimate tissue mechanical properties (stiffness and viscosity). High-resolution MRE is also used to map these properties spatially.
  • Diffusion Tensor Imaging (DTI): DTI provides information about the orientation of white matter fibers, which is crucial for understanding the anisotropic mechanical properties of the brain.
  • Anatomical MRI (T1- and T2-weighted): These sequences are used to obtain detailed anatomical information, including the geometry of the brain, the subarachnoid space, and vasculature.

The data collected from these MRI techniques are then processed to extract displacement fields, strain tensors, material properties, and anatomical features.

Datasets

The paper describes the collection of a comprehensive dataset from human volunteers, made publicly available on the Neuroimaging Tools and Resources Collaboratory (NITRC) website. This dataset includes:

  • Tagged MRI Data: Estimated displacement fields, calculated strain fields, head kinematics (angular positions, velocity, acceleration), and corresponding anatomical MR images.
  • MRE Data: Raw displacement fields, calculated strain fields, estimated shear modulus maps, head acceleration (if recorded), and corresponding anatomical MR images.

The data is collected from approximately 300 participants across a wide age range (14-80 years old), split into groups to capture age and sex differences. Processed anatomical data is registered to a standardized space (MNI-152). This dataset represents a novel benchmark for evaluating computational TBI models by providing in vivo, subject-specific biomechanical data.

Key Insights and Ideas

  • Bridging the In Vivo Gap: The core idea is to leverage non-injurious MRI measurements to overcome the ethical and practical limitations of directly studying TBI in living humans. This provides a crucial link between idealized computational models and real-world human brain responses.
  • Multi-Modal Biomechanics: The study emphasizes the importance of combining different MRI techniques (tagged MRI for dynamics, MRE for material properties, DTI for microstructure) to create a holistic picture of brain biomechanics.
  • Subject-Specific Modeling: The availability of detailed anatomical and biomechanical data for individual subjects opens the door for highly personalized computational models, which can better predict individual TBI risk.
  • Data-Driven Model Development: The paper strongly advocates for using experimental data to drive the development, calibration, and validation of computational models, rather than relying solely on assumptions or data from non-human models.

Key Results

The paper primarily focuses on describing the methodology and the dataset generated. While specific quantitative results on model performance are not presented, the key outcomes are:

  • Comprehensive Dataset: A publicly available dataset of in vivo human brain biomechanics, including deformation and material properties under controlled sub-injury loading conditions.
  • Methodological Advancements: Demonstration of the utility of tagged MRI and MRE for capturing detailed 3D brain deformation and material properties, respectively.
  • Foundation for Model Improvement: The data provides the necessary inputs (anatomy, microstructure, material properties) and validation targets (deformation fields) for developing and evaluating next-generation computational TBI models.

Limitations and Future Work

The authors acknowledge several limitations and suggest future directions:

  • Sub-Injury Levels: The current data is collected under sub-injury acceleration levels. While valuable for model development and calibration, it doesn't directly capture the complex, large-strain behavior relevant to injurious TBI events. Future work may involve exploring higher loading conditions or integrating with cadaveric data.
  • Isotropic Assumption in MRE: Standard MRE typically assumes isotropic material properties, which doesn't fully account for the anisotropic nature of white matter. Future research aims to incorporate this anisotropy into MRE analysis.
  • Limited Subject-Specific Data: While many subjects are scanned, only a small subset will undergo all types of biomechanics imaging scans, limiting the number of fully subject-specific models that can be built using all data classes.
  • Model Evaluation Metrics: Developing objective metrics to compare complex, time-resolved, tensorial simulation outputs with experimental data remains an open challenge.
  • Future Work: The authors envision broadening the types of motion and impact durations in tagged MRI experiments, exploring harmonic excitation frequencies in MRE, and developing more sophisticated computational models that can leverage this rich dataset for improved TBI prediction and prevention.

Summary updated

BibTeX
@article{bayly2021mr,
  author = {Philip V. Bayly and Ahmed Alshareef and Andrew K. Knutsen and Kshitiz Upadhyay and Ruth J. Okamoto and Aaron Carass and John A. Butman and Dzung L. Pham and Jerry L. Prince and K.T. Ramesh and Curtis L. Johnson},
  title = {{MR Imaging of Human Brain Mechanics In Vivo: New Measurements to Facilitate the Development of Computational Models of Brain Injury}},
  journal = {Annals},
  year = {2021},
  doi = {10.1007/s10439-021-02820-0},
}
Figure