Automated Neuromuscular Assessment: Machine-Learning-Based Facial Palsy Classification Using Surface...

Area of Science:

• Biomedical Engineering
• Neurology
• Machine Learning Applications

Background:

• Facial palsy (FP) significantly impacts facial muscle control, leading to asymmetry and emotional expression difficulties.
• Current assessment methods like the House-Brackmann (HB) score are subjective and examiner-dependent.
• Objective and consistent FP assessment is crucial for effective diagnosis and management.

Purpose of the Study:

• To develop and evaluate an automated system for predicting facial palsy severity using facial surface electromyography (EMG).
• To compare the performance of various machine learning models in classifying FP grades based on EMG data.
• To establish a non-invasive, objective alternative to traditional clinical scoring methods for FP.

Main Methods:

• Facial surface EMG data were collected from 58 subjects during specific facial movements (smile, eye closure, forehead raising).
• Time-domain EMG features were extracted and analyzed using nine different machine learning models.
• Model performance was assessed using accuracy, precision, recall, and F1-score metrics.

Main Results:

• Ensemble-based machine learning models, specifically random forest and decision tree ensembles, demonstrated the highest effectiveness.
• Classification accuracies for these models ranged from 81.7% to 84.8%, depending on the facial movement analyzed.
• The models reliably distinguished between different grades of facial palsy using non-invasive EMG signals.

Conclusions:

• Ensemble-based machine learning models provide a robust and reliable method for automated facial palsy grading using EMG.
• This approach offers an objective alternative to subjective clinical assessments, enhancing diagnostic consistency.
• The proposed method has the potential to improve longitudinal monitoring in both clinical and research settings for facial palsy.