Predicting Motor Imagery Performance From Resting-State EEG Using Dynamic Causal Modeling.

Area of Science:

• Neuroscience
• Brain-Computer Interfaces
• Cognitive Science

Background:

• Motor imagery-based brain-computer interfaces (MI-BCIs) enable control through imagined movements.
• Inconsistent performance and BCI-inefficiency, where users cannot generate usable brain signals, pose significant challenges.
• Understanding the neural basis of MI performance is crucial for BCI advancement.

Purpose of the Study:

• To identify resting-state network connections influencing motor imagery (MI) performance.
• To predict MI performance using these identified neural connections.
• To investigate the role of the motor network, including the dorsolateral prefrontal cortex, in MI.

Main Methods:

• Utilized a public MI dataset with resting-state fMRI and psychological questionnaires.
• Employed dynamic causal modeling (DCM) to analyze directed connectivity strengths in resting-state networks.
• Focused analysis on the motor network, specifically the supplementary motor area and dorsolateral prefrontal cortex.

Main Results:

• A significant difference in connectivity strength from the supplementary motor area to the right dorsolateral prefrontal cortex was observed between low- and high-MI performance groups.
• This specific resting-state connection was significantly stronger in the high-MI performance group.
• Resting-state connection strength positively correlated with MI-BCI performance (r=0.54, r=0.42) and was used to predict performance (r-squared=0.31).

Conclusions:

• Resting-state functional connectivity, particularly from the supplementary motor area to the right dorsolateral prefrontal cortex, is a significant predictor of MI-BCI performance.
• These findings offer potential biomarkers for BCI-inefficiency and suggest novel strategies for improving BCI system design and user training.
• Understanding pre-existing neural network properties can enhance BCI efficacy and accessibility.