Analysis of Motor Imagery-Based fNIRS-BCI for Knee Flexion and Extension

This study aimed to explore the potential of functional near-infrared spectroscopy (fNIRS) as a non-invasive technique for detecting hemodynamic responses in cortical brain activities during motor imagery (MI) tasks. Specifically, the study focused on classifying multiclass MI tasks with fNIRS data. A total of 10 healthy individuals participated in the experiment, where they were instructed to perform motor imagery tasks. A multi-channel continuous wave fNIRS system records the signals from the motor cortex for knee flexion and extension imagery tasks. Discriminant Analysis (LDA) and Support Vector Machine (SVM), comparatively computationally efficient, algorithms were used to classify the data for fNIRS-based BCI applications. The SVM classifier performed comparatively better than the LDA classifier in accuracy, with 88.8% and 89.4% for two-feature and three-feature combinations. Brain activation maps show the increased activation for flexion movement of the knee. Future research should focus on refining the classification process through feature selection, preprocessing improvements, and ensemble methods to enhance accuracy.