Predicting Mental Imagery-Based BCI Performance from Personality, Cognitive Profile and Neurophysiological Patterns.

Area of Science:

• Neuroscience
• Human-Computer Interaction
• Rehabilitation Engineering

Background:

• Mental-Imagery based Brain-Computer Interfaces (MI-BCIs) offer intuitive control via brain activity (EEG) but suffer from significant user performance variability.
• Existing research on predictors of MI-BCI control is limited, primarily focusing on motor-imagery and single-session training.

Purpose of the Study:

• To investigate predictors of MI-BCI control performance across multiple sessions.
• To explore relationships between user profiles (personality, cognitive abilities, neurophysiology) and MI-BCI control.
• To develop and validate a predictive model for MI-BCI performance.

Main Methods:

• 18 participants trained on an EEG-based MI-BCI using three mental tasks (two non-motor) over six sessions.
• Assessed relationships between BCI performance and personality, cognitive tests (including mental rotation), and neurophysiological markers.
• Developed a predictive model using psychometric data and validated it with leave-one-subject-out cross-validation.

Main Results:

• No significant correlations were found between MI-BCI performance and neurophysiological markers.
• Strong positive correlations were observed between MI-BCI performance and mental-rotation scores, indicating spatial ability is a key predictor.
• A predictive model based on psychometric questionnaires achieved high accuracy, with a mean error of less than 3 points.

Conclusions:

• User's cognitive profile, specifically spatial ability, significantly impacts MI-BCI control performance.
• Neurophysiological markers are not reliable predictors of MI-BCI control ability.
• The developed predictive model offers a reliable method for assessing and potentially improving MI-BCI user training protocols.