Effect of modeling subject-specific cortical folds on brain injury risk prediction under blunt impact loading

Abstract: Purpose: Computational head models are essential tools for studying the risk of mild traumatic brain injury (mTBI) under different activities and across populations. However, different computational models incorporate varied levels of anatomical details, such as cortical folds. In this study, we aim to determine the effect of modeling cortical folds on mTBI risk assessment. Methods: We compared the gyrencephalic (with cortical folds) and lissencephalic (without cortical folds) FE models of 18 subjects aged 9 - 18 years, under a rotational head acceleration event. A rotational acceleration of 10 krad/s$^2$ and 10 ms duration was simulated about each principal head axis. We analyzed different mTBI injury metrics, including maximum principal strain (MPS95), maximum principal strain rate (MPSR95), and cumulative strain damage measure (CSDM15), for the whole brain as well as for specific regions of interest (ROIs). Results: Modeling cortical folds consistently predicted higher injury metrics across all individuals and rotational direction, with the bias (mean $\pm$ std. dev.) of $-15.1 \pm 6.5\%$ in MPS95, $-12.9 \pm 5.6\%$ in MPSR95, and $-8.8 \pm 11.09\%$ in CSDM15. We also find that the regions of high strain concentrations vary significantly between the two models, with the DICE metric on peak MPS ranging between $0.07-0.43$ and DICE on CSDM15 ranging between $0.42-0.70$. Modeling cortical folds also affects injury metrics in all ROIs, even the ones that remain geometrically unaltered in the two model types, such as the corpus callosum, cerebellum, and brain stem. Conclusions: The study finds that modeling cortical folds significantly alters the region of high brain deformations and the mTBI risk under head rotations.