A Deep Learning Model for Automated Classification of Intraoperative Continuous EMG.

Area of Science:

• Neuroscience
• Artificial Intelligence
• Medical Technology

Background:

• Intraoperative neurophysiological monitoring (IONM) uses electrophysiological methods to safeguard nerves during high-risk surgeries.
• Current IONM is hindered by communication delays, subjective signal interpretation, and inter-rater variability.

Purpose of the Study:

• To develop an automated system for classifying electromyogram (EMG) waveforms during IONM.
• To address limitations in current IONM practices through advanced signal processing and machine learning.

Main Methods:

• A hybrid Convolutional Neural Network (CNN) and Long Short-Term Memory (LSTM) model was developed for EMG waveform classification.
• A data normalization preprocessing pipeline was implemented to manage multi-subject data.
• Model robustness was evaluated under various artifact processing techniques.

Main Results:

• The proposed CNN-LSTM model achieved 89.54% accuracy and 94.23% sensitivity in cross-patient evaluation.
• This performance surpassed several benchmark modeling methods.
• The model effectively captured complex EMG patterns amidst electrical noise and movement artifacts.

Conclusions:

• The CNN-LSTM model demonstrates significant potential for automated continuous EMG classification in IONM.
• This technology can enhance surgical safety by minimizing cognitive load and inter-rater variability for surgeons.