Influence of Strain Post-Processing on Brain Injury Prediction

Finite element head models are a tool to better understand brain injury mechanisms. Many of the models use strain as output but with different percentile values such as 100 th , 95 th , 90 th , and 50 th percentiles. Some use the element value, whereas other use the nodal average value for the element. Little is known how strain post-processing is affecting the injury predictions and evaluation of different prevention systems. The objective of this study was to evaluate the influence of strain output on injury prediction and ranking. Two models with different mesh densities were evaluated (KTH Royal Institute of Technology head model and the Total Human Models for Safety (THUMS)). Pulses from reconstructions of American football impacts with and without a diagnosis of mild traumatic brain injury were applied to the models. The value for 100 th , 99 th , 95 th , 90 th , and 50 th percentile for element and nodal averaged element strain was evaluated based on peak values, injury risk functions, injury predictability, correlation in ranking, and linear correlation. The injury risk functions were affected by the post-processing of the strain, especially the 100 th percentile element value stood out. Meanwhile, the area under the curve (AUC) value was less affected, as well as the correlation in ranking (Kendall’s tau 0.71-1.00) and the linear correlation (Pearson’s r 2 0.72-1.00). With the results presented in this study, it is important to stress that the same post-processed strain should be used for injury predictions as the one used to develop the risk function.