Classification of left and right foot kinaesthetic motor imagery using common spatial pattern

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Brain-computer interface (BCI) systems commonly use Common Spatial Pattern (CSP) for upper-limb motor imagery but struggle with foot imagery due to its location.
• Classifying left versus right foot Kinaesthetic Motor Imagery (KMI) presents challenges for existing BCI feature extraction methods.

Purpose of the Study:

• To develop and evaluate novel methods for classifying left and right foot KMI using EEG.
• To optimize subject-specific band selection for improved feature extraction in BCI.

Main Methods:

• Applied CSP and Filter Bank Common Spatial Pattern (FBCSP) for mu and beta rhythm feature extraction during foot KMI.
• Utilized logistic regression (Logreg) and linear discriminant analysis (LDA) for classification, validated with 10-fold cross-validation.
• Compared four paradigms: CSP LDA, CSP Logreg, FBCSP LDA, and FBCSP Logreg.

Main Results:

• All BCI paradigms exceeded the statistical chance level of 60.0% accuracy.
• FBCSP LDA demonstrated superior performance with an average accuracy of 70.28% ± 4.23 and a kappa score of 0.41.
• The FBCSP LDA paradigm achieved a maximum classification accuracy of 77.5% for discriminating between left and right foot KMI in single-trial analysis.

Conclusions:

• The proposed CSP and FBCSP methods offer a viable approach for classifying left versus right foot KMI.
• These findings support the potential application of these BCI paradigms for controlling robotic foot devices or neuroprostheses.