Concussion case classification based on brain strain waveforms using dynamic time warping and cluster analysis

Abstract

In sports, concussion prevention measures require the development of concussion evaluation criteria for improved head protective gear or early intervention through impact measurement in the field. The peak value of the brain strain has been used as a brain strain metric to evaluate the risk of brain injury. However, it extracts the instantaneous moment of strain history experienced by the brain tissue; therefore, it lacks the characteristics of the entire strain waveform experienced by the brain tissue during a head impact. This study developed a new method for classifying concussion cases based on the overall characteristics of the brain strain waveform. First, 53 head collision cases tagged with the concussion status in American football were simulated using a finite element head model, and the brain strain waveforms in four regions (right brain, left brain, brainstem, and cerebellum) were obtained. By applying the dynamic time-warping distance and clustering method to the brain strain waveforms in each case, a concussion classification method that considers the overall characteristics of the waveforms was constructed. The classification evaluation results showed that the concussion classification method had a higher performance than the conventional method using the peak value of brain strain. Brain strain waveforms in concussion cases have multiple high peaks, and it is important to understand the characteristics of the entire brain strain waveform, not just the instantaneous peak value, to classify concussion cases.

Keywords

• concussion
• sports injury
• brain strain waveform
• finite element method (fem)
• clustering analysis