Post-concussive mTBI in Student Athletes: MRI Features and Machine Learning

Area of Science:

• Neuroimaging
• Radiomics
• Machine Learning

Background:

• Post-concussive syndrome (PCS) affects student athletes after mild traumatic brain injury (mTBI).
• Identifying mTBI and its impact on athletes requires advanced diagnostic tools.

Purpose of the Study:

• To identify radiomics features from structural MRI (MPRAGE) and Diffusion Tensor Imaging (DTI) associated with mTBI in student athletes with PCS.
• To develop a machine learning model for predicting mTBI history in this population.

Main Methods:

• 122 student athletes with PCS and 27 controls underwent MPRAGE and DTI MRI scans.
• Radiomic features were extracted from white and gray matter regions.
• Five Support Vector Machines were trained and validated; top models were ensembled for prediction.

Main Results:

• The ensembled model achieved 80% sensitivity and 74% specificity in distinguishing PCS subjects from controls.
• White matter radiomics features showed a strong association with mTBI.
• The predictive index correlated significantly with the number of concussions and time from injury.

Conclusions:

• MRI-derived radiomics features are associated with mTBI history in student athletes with PCS.
• A predictive machine learning model for mTBI was successfully developed using these features.