J16 - Characterizing the Relationship between Cranial Morphometry and Brain Material Properties using MR Imaging and Elastography

Introduction: : Traumatic brain injury (TBI) poses significant global health risks, prompting the need for predictive computational models to assess injury risk and improve safety in engineering design [1]. Subject-specific imaging-based computational models have shown that the amount of deformation depends on individual geometric and material properties, like brain size and stiffness [2]. However, current computational models frequently rely on the geometry of a 50th-percentile male head, as subject-specific magnetic resonance (MR) imaging and elastography data is costly and not always available [1,2]. As a result, there is a need to develop methods to scale subject-specific computational models without relying on MR data [3]. This study aims to advance subject-specific TBI modeling by investigating correlations between subject anthropometry (age, sex, head dimensions) and brain morphometry from magnetic resonance (MR) images [Aim 1] and between head anthropometry and brain material properties derived from MR elastography [Aim 2] using statistical models.

Materials and

Methods: : Neuroimaging data, including MR structural imaging and elastography, were sourced from a recently acquired open-source dataset [3]. The structural imaging cohort contained 159 subjects (76 males, 83 females), aged 14-75, with T1- and T2-weighted images and associated data (age, weight, height). A standard protocol was developed to obtain manual measurements of the head size and shape dimensions in ITK-Snap. Brain dimensions were extracted from a binary brain mask using a custom code in MATLAB R2024b, and brain volume was obtained from a deep-learning segmentation [4]. Whole-brain MR elastography data, collected at 30, 50, and 70 Hz frequencies, was analyzed from a subset of 85 subjects (37 males, 48 females). Median and 95th-percentile material properties, including shear stiffness, damping ratio, storage modulus, and loss modulus, were extracted. Mathematical models (simple/stepwise linear regression) were developed to predict brain morphometry and material properties from subject anthropometry. Principal component (PC) analysis was performed on a subset of 89 subjects (37 males, 54 females) to reduce dimensionality and identify brain shape variation. The resulting PCs were correlated with strain measures to evaluate how variability in brain morphometry influences the strain response.

Results, Conclusions, and Discussions:: The correlations between respective head and brain dimensions showed predictive potential, although not all variation was explained. Simple linear regression (Figure 1) indicated that head dimensions are moderate predictors of respective brain length, width, and depth (R; 0.75, 0.88, 0.76, respectively), while stepwise linear regression lowered prediction error, modeling brain dimensions as specific linear combinations of head length, width, or depth (p < 0.05). Stepwise linear regression improved prediction accuracy, suggesting brain dimensions are influenced by combinations of head length, width, and depth. Gender-specific analysis revealed that males presented an 11.34 % larger average brain volume (1316 ± 110 cm³) compared to females (1182 ± 114 cm³), with larger head and brain dimensions on average, confirming existing literature [5]. Age-related trends showed a linear decline in brain volume with age of approximately 2.9 cm³ per year. This emphasized the need to scale computational injury prediction models to account for subject-specific variation. Analysis of the correlations between head size and the MR elastography dataset showed no significant relationship. Brain mechanical properties (shear stiffness, storage modulus, loss modulus) exhibited a significant relationship with subject age and ventricular volume (R² > 0.5), but no other head or brain dimensions. Brain shape analysis revealed that five PCs (sagittal tilt, axial tilt, coronal tilt, brain volume, and a width-to-length metric) accounted for 95 % of the variance in brain morphometry (Figure 2). The relationship between these components and brain strain response is currently under investigation, with initial results showing higher correlations to those PCs than standard dimensions.

Acknowledgements and/or References (Optional):: [1] Wu S, Zhao W, Wu Z, McAllister T, Hu J, Ji S. Approximating subject-specific brain injury models via scaling based on head-brain morphological relationships. Biomech Model Mechanobiol. 2023; 22:159-175.
[2] Knutsen A, Gomez A, Gangolli M, Wang W, Chan D, Lu Y, Christoforou E, Prince J, Bayly P, Butman J, Pham D. In vivo estimates of axonal stretch and 3D brain deformation during mild head impact. Brain Multiphysics. 2020; 1:100015.
[3] Bayly P, Alshareef A, Knutsen A, Upadhyay K, Okamoto R, Carass A, Butman J, Pham D, Prince J, Ramesh K, Johnson C. MR imaging of human brain mechanics in vivo: new measurements to facilitate the development of computational models of brain injury. Ann Biomed Eng. 2021; 49:2677-2692.
[4] Huo Y, Plassard A, Carass A, Resnick S, Pham D, Prince J, Landman B. Consistent cortical reconstruction and multi-atlas brain segmentation. Neuroimage. 2016; 138:197-210.
[5] Ruigrok A, Salimi-Khorshidi G, Lai M, Baron-Cohen S, Lombardo M, Tait R, Suckling J. A meta-analysis of sex differences in human brain structure. Neurosci Biobehav Rev. 2014; 39:34-50.