Biomechanically informed improvements to the brain–skull interface in an atlas-based brain model

Introduction Finite element (FE) models are powerful tools for injury analysis and prediction. Their performance should be assessed with a standardized system to validate their use before employing them for injury prediction. It is necessary to develop an FE model that is able to reproduce the local response of a wide range of experimental conditions for the intended use. Because there is no standard for quantifying model performance, different validated brain models can produce varying regional pressure and strain responses, which motivates the use of a standard validation protocol, as used in the current study. Methods In the current study, biofidelity improvements were made to an existing atlas-based brain FE model developed from the International Consortium for Brain Mapping (ICBM) brain atlas. The brain-skull interface was enhanced by decoupling the current skull and cerebrospinal fluid (CSF) parts and implementing a smooth skull geometry and a sliding with no separation contact at the skull interface. This new model is an enhancement of the atlas-based brain model (ABM) and will be referred to as ABMv4. The new, decoupled model was tested in various experimental impacts and its performance against validation data assessed using CORA (CORrelation and Analysis), an objective rating method. Results A smooth surface at the skull interface was implemented to create the ABMv4. The ABMv4 was then validated against 23 experimental cadaver impacts - sixteen ICP tests, nine strain tests, and seventeen brain displacement tests. The improved model achieved CORA ratings of 0.733, 0.563, 0.846, and 0.605 for ICP, brain strain, revised brain strain, and brain displacement validation tests, respectively. Discussion The ABMv4 performs well in all types of validation tests and achieves a good biofidelity rating. Additionally, the ABMv4 results in increased relative brain-skull motion (average of 0.54 mm for ABMv2 compared to average of 1.06 mm for ABMv4) and reduced brain strain on the surface of the model, with the surface strain being on average 61.7% of the surface strain value predicted for the ABMv2.