Abstract
One of the most significant challenges in the application of brain-computer interfaces (BCI) is the large performance variation, which often occurs over time or across users. Recent evidence suggests that the physiological states may explain this performance variation in BCI, however, the underlying neurophysiological mechanism is unclear. In this study, we conducted a seven-session motor-imagery (MI) experiment on 20 healthy subjects to investigate the neurophysiological mechanism on the performance variation. The classification accuracy was calculated offline by common spatial pattern (CSP) and support vector machine (SVM) algorithms to measure the MI performance of each subject and session. Relative Power (RP) values from different rhythms and task stages were used to reflect the physiological states and their correlation with the BCI performance was investigated. Results showed that the alpha band RP from the supplementary motor area (SMA) within a few seconds before MI was positively correlated with performance. Besides, the changes of RP between task and pre-task stage from theta, alpha, and gamma band were also found to be correlated with performance both across time and subjects. These findings reveal a neurophysiological manifestation of the performance variations, and would further provide a way to improve the BCI performance.
Introduction
Brain-Computer Interfaces (BCIs) based on Motor Imagery (MI)—where a user controls a computer by simply imagining movement—hold massive potential for stroke rehabilitation. However, they face a major "usability gap." Roughly 30% of users are considered "BCI illiterates," meaning they struggle to reach proficiency even with training. Furthermore, a user’s performance can swing wildly from one day to the next.
The challenge is that we don't fully understand the neurophysiological "why" behind this variation. If we can identify stable biomarkers that correlate with high performance, we can design better training protocols or adaptive systems that "meet the user where they are" based on their current physiological state.
Methods
The researchers conducted a longitudinal study with 20 healthy, right-handed individuals (11 males, mean age 23.2) over two weeks, totaling seven sessions per person. Each session consisted of 6 runs, with 40 trials per run.
The Task and Signal Processing
Subjects performed four MI tasks: imagining left-hand movement, right-hand movement, both feet movement, or an "idle" state. To clean the noisy EEG data, Independent Component Analysis (ICA) and a sophisticated hybrid method called EEMD-CCA (Ensemble Empirical Mode Decomposition and Canonical Correlation Analysis) were used to strip away muscle artifacts.
Feature Extraction and Classification
Instead of looking at raw power, the study calculated Relative Power (RP)—the ratio of power in a specific frequency band (Theta, Alpha, Beta, Gamma) to the total power. This helps normalize data across different individuals. They used the standard Common Spatial Pattern (CSP) algorithm for spatial filtering and a Support Vector Machine (SVM) to decode the brain signals into commands.
Key Results
The study found significant variation in performance across the 20 subjects. The average three-class accuracy was 50.43% ± 10.49% (well above the 33.3% chance level), while binary classification (dominant hand vs. idle) reached 68.95% ± 10.54%.
Neurophysiological Correlations
- The "Ready State" Predictor: Alpha band RP in the Supplementary Motor Area (SMA) measured just before the task began was a strong positive predictor of success ($r = 0.546$ for binary accuracy). Essentially, the brain's "resting" state in the motor planning region determines how effectively it can transition into an active imagery state.
- Dynamic Range: High performers showed much larger changes ($\Delta RP$) when moving from rest to task. Specifically, high performance was linked to a sharper decrease in alpha power (Event-Related Desynchronization) and a sharper increase in theta and gamma power.
- Stability Across Time: These markers (especially in the alpha and gamma bands) remained significant predictors even as a user's performance fluctuated over the two-week period.
Discussion and Future Work
The Supplementary Motor Area emerged as the most critical region for these correlations, highlighting its role in the "mental rehearsal" of movement. Interestingly, the beta band—usually a staple of motor imagery—did not show significant correlations, possibly due to the specific RP calculation method used.
While the results are promising in healthy subjects, the next step is to validate these biomarkers in stroke patients, whose brain signals may be significantly altered by neural damage. Ultimately, this research paves the way for "neurofeedback" systems that could assess a user's physiological state before beginning a BCI training session, potentially closing the "usability gap" for BCI technology.
Summary updated