Sports Concussion And Mild Traumatic Brain Injury Assessment Is Enhanced With Portable, Non-invasive Biosensor Arrays

Abstract

PURPOSE: To determine the degree that sports concussion may be identified by predictive multivariate models including neuropsychological testing data, single-lead electroenceplographic (EEG) data, and other modes of biosensor data. METHODS: Collegiate athletes who suffered traumatic brain injury (concussion; N=30) while participating in varsity sports were assessed in a longitudinal study design during the symptomatic and asymptomatic phase of recovery. These data were compared to an equal number of non-injured, age matched, comparator subjects. Each participant was tested using a novel single-lead EEG device during a series of cognitive and behavioral tasks, including eyes closed (EC) and eyes open (EO) resting task and a five word immediate recall task under IRB supervision. Behavioral measures (e.g. error-rate/task duration) and EEG spectral and wavelet features were extracted from EC/EO baseline and neuropsychological test epochs. Univariate and multivariate logistic regression, linear discriminant analysis and random forest models were compared for the ability to predict a concussive event as well as to develop signatures of Return-to-Learn and Return-to-Play decisions. RESULTS: EEG spectral density power features (e.g. relative beta power), relative to overall power, had an overall predictive accuracy of 65% (p<0.004). Additionally, total read times alone were 62% accurate (p<0.015) in a neuro-ophthalmologic saccade test. In a multivariate model, these two modes were found to be 70% accurate (p<0.006). However, inclusion of covariates increased discriminant power to 76% accurate at differentiating concussed from control subjects (p<0.003). CONCLUSIONS: Neuropsychological testing benefits by the addition of other objective biosensor modalities in the predictive performance of concussion / TBI testing for both the identification and management in Return-to-Learn and Return-to-Play decision making. These results suggest the use of multivariate signatures for concussion/TBI testing. Nonetheless, these preliminary findings must be independently replicated, validated, and expanded in sample size across multiple clinical settings such as team sports and emergency department use.