Automatic Grading of Stroke Symptoms for Rapid Assessment Using Optimized Machine Learning and 4-Limb Kinematics:...

Area of Science:

• Neurology
• Biomedical Engineering
• Machine Learning

Background:

• Subtle motor deficits in neurological diseases require timely treatment, but objective assessment by non-specialists is challenging.
• Current clinical grading relies on subjective scales like the National Institutes of Health Stroke Scale (NIHSS) and Medical Research Council (MRC) scores.
• Objective, real-time motor function assessment is crucial for consistent diagnosis and rapid medical intervention.

Purpose of the Study:

• To develop an autonomous grading system for stroke patients to objectively assess motor weakness.
• To evaluate the system's feasibility in grading National Institutes of Health Stroke Scale (NIHSS) and Medical Research Council (MRC) scores for all four limbs.

Main Methods:

• An automated system was created using wearable inertial sensors for measuring subtle limb weaknesses and machine learning algorithms for grading.
• Data from 60 stroke patients were collected, and a synthetic minority oversampling technique (SMOTE) was used to generate 240 training instances for imbalanced data.
• Ensemble and Support Vector Machine (SVM) algorithms were trained and optimized using Bayes optimization and 5-fold cross-validation.

Main Results:

• The automated system achieved an 83.3% accuracy and 0.912 AUC for National Institutes of Health Stroke Scale (NIHSS) grading using an ensemble algorithm.
• Support Vector Machine (SVM) achieved 80.0% accuracy and 0.860 AUC for NIHSS grading.
• Auto-MRC grading demonstrated accuracies of 76.7% (SVM) and 78.3% (ensemble), with mean AUCs of 0.870 and 0.877, respectively.

Conclusions:

• The developed automated grading system quantifies proximal weakness in real-time, demonstrating feasibility for remote stroke monitoring.
• The system enables consistent, instant assessment and objective scoring (auto-MRC and auto-NIHSS), facilitating expedited hospital dispatches and treatment initiation.
• This technology can improve prehospital and hospital care coordination through objective, shared patient data.