Work

Machine learning classification of acute concussion using data obtained from neurological tests delivered through virtual reality

Year: 2026
Type: article
Study design: Observational Study
Abstract: Concussion is a traumatic brain injury with a subtle and divergent clinical presentation, which may account for why a gold standard for detection has yet to be found. This study describes the initial machine learning (ML) classification results of a project using virtual reality (VR) tests to detect concussion. Participants aged 18–40 with (cases = 29) and without (controls = 697) an acute concussion were enrolled. Self-reported and sensor-based data from the VR headset were collected. Altogether, 175 features were constructed from VR sensor-based data. Additionally, age, sex, varsity status, handedness, and year of education were used as independent features. Using two classifiers, logistic regression with an L1 penalty (LASSO) and a gradient boosting classifier (HGB), five models that vary in VR sensor, symptom, and participant characteristic features were developed. Quality metrics include area under the receiver operating characteristic curve (AUC) and balanced accuracy. The full model showed the highest discriminability with average AUC of 0.86 for LASSO (95%CI 0.66–0.98) and 0.85 for HGB (95% CI 0.67–0.97). The average balanced accuracy for this model using LASSO was 0.77 (95% CI 0.56–0.91) and was 0.69 (95% CI 0.49–0.90) for HGB.
Language: English
Field: Medicine
Open Access status: gold
Estimated APC paid: $2,690
OpenAlex ID: W7225911807
FWCI:
Cites: 35
Cited by:
Related to: 10