Experimental Analysis of Automatic Discrimination Performance Between Simulated Bruxism and Non-Bruxism Under...

Area of Science:

• Biomedical Engineering
• Machine Learning Applications
• Neuromuscular Monitoring

Background:

• Bruxism, characterized by teeth grinding or clenching, can be associated with tooth contact or occur without it.
• Accurate classification of different bruxism movements is crucial for diagnosis and management.
• Electromyography (EMG) and sound data offer potential biomarkers for detecting bruxism events.

Purpose of the Study:

• To evaluate the efficacy of machine learning algorithms in automatically classifying electromyography (EMG) data.
• To differentiate between bruxism simulated movement with tooth contact (BMwTC), bruxism simulated movement without tooth contact (BMwoTC), and non-bruxism movement (non-BM).
• To assess the performance of single- and multi-stream Hidden Markov Models (HMMs) for automated bruxism classification.

Main Methods:

• Twelve healthy participants performed simulated bruxism movements (BMwTC, BMwoTC, non-BM).
• EMG data were recorded from masseter, infrahyoid, inframandibular, and chin muscles, along with adjacent sound data.
• Single- and multi-stream HMMs were employed for classification using a leave-one-out cross-validation approach.

Main Results:

• The single-stream HMM using masseter EMG achieved significantly higher discrimination accuracy (p < 0.05).
• While the multi-stream model showed improved accuracy, the difference was not statistically significant.
• Classification accuracy for BMwoTC was notably below 0.5, indicating challenges in distinguishing this specific movement.

Conclusions:

• A machine learning-based system effectively discriminates BMwTC from non-BM using masseter EMG signals.
• The findings suggest the potential of automated EMG analysis for bruxism detection.
• Further research may be needed to improve the classification accuracy for BMwoTC events.