Functional outcome prediction after traumatic cervical spinal cord injury using ensemble machine learning: a...

Area of Science:

• Neurology
• Medical Imaging
• Machine Learning

Background:

• Traumatic cervical spinal cord injury (TCSCI) frequently results in significant neurological deficits.
• Predicting functional recovery is crucial for effective patient management and rehabilitation strategies.

Purpose of the Study:

• To create an ensemble learning model for predicting one-year neurological and functional outcomes in TCSCI patients.
• The model integrates baseline clinical data, neurological assessments, and cervical MRI features.

Main Methods:

• A two-layer Stacking ensemble model was developed using data from 410 TCSCI patients.
• Predictors included demographic, clinical, and radiologic features, with outcomes like AIS grade and motor scores.
• SHapley Additive exPlanations (SHAP) analysis was used for model interpretability.

Main Results:

• The model demonstrated high accuracy, achieving AUC ≥ 0.85 for AIS grades.
• Excellent R² values (≥ 0.986) were observed for Upper Extremity Motor Score (UEMS), Lower Extremity Motor Score (LEMS), Total Motor Score (TMS), and Spinal Cord Independence Measure III (SCIM III).
• SHAP analysis highlighted baseline UEMS and AIS grade as key predictors.

Conclusions:

• The externally validated model accurately predicts one-year functional recovery in TCSCI patients.
• This tool can assist in early prognosis and personalized rehabilitation planning.