Automatic generation and validation of patient-specific finite element head models suitable for crashworthiness analysis

Abstract

A method to automatically generate finite element (FE) head models is presented in this paper. Individual variation in geometry of the head should be taken into consideration in future injury-prediction research. To avoid inter- and intra-operator variation due to manual segmentation, a robust and accurate algorithm is suggested. The current approach utilises expectation maximisation classification and skull stripping. The whole process from geometry extraction to model generation is converted into an automatic scheme. The models that are generated from the proposed method are validated in terms of segmentation accuracy, element quality and injury-prediction ability. The segmentations of the white matter and grey matter are about 90% accurate and the models have good element quality, with 94% of the elements having a Jacobian above 0.5. Using the experimental data from post-mortem human subject heads, nodal displacements were compared with the data collected from the simulations with the FE head models. The results are promising, indicating that the proposed method is good enough to generate patient-specific model for brain injury prediction. Further improvement can be made in terms of geometry accuracy and element quality.