Riemannian geometry-based metrics to measure and reinforce user performance changes during brain-computer interface...

Area of Science:

• Neuroscience
• Biomedical Engineering
• Machine Learning

Background:

• Brain-computer interfaces (BCI) show promise but face limitations due to BCI inefficiency, where users struggle to generate control signals.
• Effective user training protocols are crucial to enhance BCI usability and reduce inefficiency.
• Performance assessment and feedback mechanisms are key components for successful BCI skill acquisition.

Purpose of the Study:

• To develop and evaluate novel trial-wise metrics based on Riemannian geometry for assessing user performance in BCI.
• To adapt existing Riemannian geometry metrics (classDistinct and classStability) for real-time feedback during BCI training.
• To compare the effectiveness of these new metrics against conventional classifier feedback in reflecting user performance trends.

Main Methods:

• Introduced three trial-wise adaptations of Riemannian geometry metrics: running, sliding window, and weighted average.
• Calculated classDistinct (class separability) and classStability (within-class consistency) for each trial.
• Evaluated metrics using simulated and real sensorimotor rhythm-BCI data, comparing them with standard classifier feedback.

Main Results:

• The sliding window and weighted average variants of the proposed Riemannian geometry metrics demonstrated higher accuracy in reflecting performance changes during BCI sessions.
• These novel metrics showed better discrimination of user performance trends compared to conventional classifier feedback.
• The trial-wise metrics effectively captured performance variations throughout BCI usage.

Conclusions:

• Trial-wise Riemannian geometry-based metrics offer a viable approach for evaluating and monitoring user performance during BCI training.
• The sliding window and weighted average adaptations show particular promise for real-time feedback.
• Further research is recommended to explore optimal presentation methods for these metrics in user training protocols to enhance BCI adoption.