A Novel Real-time Phase Prediction Network in EEG Rhythm.

Area of Science:

• Neuroscience
• Biomedical Engineering
• Machine Learning

Background:

• Closed-loop neuromodulation utilizes electroencephalography (EEG) phase to monitor brain states and optimize stimulation.
• Non-stationary EEG signals present challenges for accurate phase-based prediction, leading to variances and reduced accuracy.
• Existing prediction methods struggle with the dynamic nature of EEG signals.

Purpose of the Study:

• To develop a novel machine learning model for precise EEG phase prediction.
• To enhance the accuracy and reduce variance in real-time brain state assessment for neuromodulation.
• To introduce the EEG phase prediction network (EPN) for capturing subject-specific rhythm patterns.

Main Methods:

• Proposed a machine learning-based network named EEG phase prediction network (EPN).
• EPN maps instantaneous phase directly from narrow-band EEG data by capturing rhythm distribution patterns.
• Validated EPN performance using pre-recorded, simulated, and real-time EEG data.

Main Results:

• EPN demonstrated superior performance compared to state-of-the-art models.
• Achieved the lowest variance and highest accuracy in EEG phase prediction.
• Successfully captured subject-specific EEG rhythm distribution patterns.

Conclusions:

• The EPN model offers a significant advancement in EEG phase prediction accuracy and reliability.
• This model is poised to broaden applications in EEG phase-based closed-loop neuromodulation.
• EPN provides a robust solution for real-time brain state monitoring in neuromodulation therapies.