Brain injury metrics and their risk functions in frontal automotive collisions.

Abstract

Objectives: The objective of this study was to develop abbreviated injury scale (AIS) 1, AIS2, AIS3 and AIS4+ injury risk functions (IRFs) for traumatic brain injuries (TBIs) as estimated by the rotational kinematics of the head, in accordance with AIS1998. The effectiveness of the IRFs was investigated by comparisons with real-world accident data of frontal crash configurations. In addition, links of the IRFs developed in accordance with AIS1998 to other AIS versions were discussed.

Methods: AIS1, AIS2, AIS3 and AIS4+ IRFs based on finite element analysis (FEA)-based metrics in this study were developed using a TBI database used for developing mild TBI (concussion) and severe TBI (diffuse axonal injury (DAI) and intracerebral hemorrhage (ICH)) IRFs in our previous study. The TBI database includes head kinematics, clinical outcomes, and FEA-based metrics such as maximum principal strain (MPS) obtained from reconstructions using harmonized species-specific finite element (FE) brain models. In this study, TBI severities in the TBI database were reclassified in accordance with AIS1998 to evaluate IRFs in comparison with field accident data for application to automotive safety. IRFs based on kinematics-based metrics were developed by transforming FEA-based IRFs via linear regression models between the FEA-based and kinematics-based metrics. The FEA-based and kinematics-based IRFs were evaluated by comparing TBI risk predictions using frontal crash test data with real-world TBI rates in similar crash configurations.

Results: The MPS95 IRFs exhibited better quality (lower quality index (QI) values) and better goodness of fit with the TBI database (lower AIC value) among the FEA-based IRFs. Kinematics-based metrics exhibited the greatest coefficients of determination (R²) with MPS95. The accident data evaluation demonstrated that the MPS95 IRFs and kinematics-based IRFs derived from the MPS95 IRFs generally overpredicted most frontal crash configurations, with the full engagement conditions tending to have smaller errors and the oblique crash conditions having the largest overprediction.

Conclusions: The TBI risks predicted by the MPS95 IRFs and kinematics-based IRFs derived from the MPS95 IRFs were relatively more aligned with the real-world TBI rates for drivers in the full engagement crash configuration. However, further investigations are needed to minimize the gap between predicted TBI risks and real-world TIB rates. In addition, AIS coding of TBIs has changed through version upgrades, especially for concussion. This change in AIS coding has affected IRFs for AIS1 and AIS2. Further revisions of TBI IRFs will be required in the future if the AIS definitions change.