Prediction of Event Related Potential Speller Performance Using Resting-State EEG

Area of Science:

• Neuroscience
• Biomedical Engineering
• Rehabilitation Technology

Background:

• Event-related potential (ERP) spellers offer communication and device control for individuals with severe motor impairments.
• Challenges include inconsistent user performance and a lack of understanding of individual ERP capabilities (ERP-illiteracy).
• Predicting speller performance before use is essential for efficient application and user selection.

Purpose of the Study:

• To identify resting-state electroencephalography (EEG) features that correlate with ERP speller performance.
• To investigate the predictive power of spectral power and functional connectivity in specific brain regions and frequency bands.
• To develop a method for predicting ERP speller classification accuracy using pre-speller EEG data.

Main Methods:

• Analysis of resting-state EEG data from participants prior to ERP speller tasks.
• Calculation of spectral power across four brain regions and five frequency bands (delta, theta, alpha, beta, gamma).
• Assessment of functional connectivity within and between brain regions across the same frequency bands.

Main Results:

• Significant correlations were found between frontal delta power and ERP speller performance.
• Functional connectivity in delta, alpha, and gamma bands also showed significant correlations with speller accuracy.
• A predictive model using these resting-state EEG features successfully estimated ERP speller performance.

Conclusions:

• Resting-state EEG spectral power and functional connectivity are viable biomarkers for predicting ERP speller performance.
• These findings can aid in identifying users likely to succeed with ERP spellers, addressing ERP-illiteracy.
• This approach may facilitate personalized BCI (brain-computer interface) selection and optimize assistive technology for patients.