Abstract
This article provides a formal correction to the study "Predicting Mental Imagery-Based BCI Performance from Personality, Cognitive Profile and Neurophysiological Patterns," originally published in PLOS ONE (DOI: 10.1371/journal.pone.0143962).
Introduction
Mental Imagery-based Brain-Computer Interfaces (MI-BCIs) enable users to control external devices through imagined movements. Despite their potential, approximately 15-30% of users struggle to achieve proficiency, a challenge often termed "BCI illiteracy." Identifying the psychological and neurophysiological traits that predict success is critical for improving training protocols. This correction addresses a data entry error regarding "Abstractedness" scores in the original study, providing a revised and more robust predictive model for BCI performance.
Methodology and Data
The researchers employed a stepwise linear regression to identify the most significant predictors of BCI performance among 18 subjects. To ensure model generalizability and prevent overfitting, a leave-one-subject-out (LOSO) cross-validation process was used. This involved generating 17 distinct models, each excluding one participant, to test predictive accuracy on the held-out data.
The dataset included:
- BCI Performance: Accuracy scores from mental imagery tasks.
- Psychometric Profiles: 16PF-5 personality factors (e.g., self-reliance, tension, apprehension).
- Cognitive Metrics: Mental Rotation Test (spatial ability) and the Corsi block-tapping task (memory span).
- Learning Styles: Dimensions such as Visual/Verbal and Active/Reflective.
Key Insights
The revised analysis yielded a more stable model explaining 87.8% of the variance in BCI performance ($R^2_{adj} = 0.878, p < 0.001$).
The Spatial Connection
Mental Rotation emerged as the most consistent predictor of BCI success. Users with higher spatial reasoning abilities—specifically the ability to mentally manipulate 3D objects—demonstrated significantly better MI-BCI performance. This suggests a strong link between the neural pathways for spatial reasoning and motor imagery.
The Impact of Anxiety
While the original model highlighted "Tension," the corrected model identifies "Apprehension" as a key predictor. Both are sub-dimensions of the broader "Anxiety" factor in the 16PF-5. High levels of anxiety consistently correlate with diminished BCI performance, identifying it as a significant barrier to mastery.
Model Stability and Refinement
The correction improved the scientific rigor of the findings. The new model proved more stable, with 13 out of 17 cross-validation iterations yielding the same factors as the global model. Interestingly, including memory span (Corsi test) was found to impair predictive power, leading to its exclusion from the final model.
Results and Limitations
The final revised model identifies four primary predictors: Mental Rotation, Self-Reliance, Apprehension, and the Visual/Verbal learning style. The model achieved a mean absolute error of only 1.00 point, with 12 out of 17 subjects' performance falling within the predicted confidence intervals. Consequently, the authors state that the "Model #2" from the original publication should be disregarded in favor of this more stable revised model.
However, limitations remain. The small sample size (n=18) increases the risk of overfitting, as evidenced by some cross-validation steps explaining nearly 100% of the variance. Additionally, the model failed to accurately predict performance for 5 subjects, often due to overly narrow confidence intervals from training data overfitting.
Conclusion
The corrected model provides a more reliable framework for understanding the human factors behind BCI success. Future research will focus on validating these predictors in larger, more diverse populations to develop standardized BCI aptitude assessments.
Summary updated