Investigating the informative brain region in multiclass electroencephalography and near infrared spectroscopy based...

In recent years, various brain imaging techniques have been used as input signals for brain-computer interface (BCI) systems. Electroencephalography (EEG) and near-infrared spectroscopy (NIRS) are two prominent techniques in this field, each with its own advantages and limitations. As a result, there is a growing tendency to integrate these methods in a hybrid within BCI systems. The primary aim of this study is to identify highly functional brain regions within an EEG + NIRS-based BCI system. To achieve this, the research focused on identifying EEG electrodes positioned in different brain lobes and then investigating the functionality of each lobe. The methodology involved segmenting the EEG + NIRS dataset into 2.4 s time windows, and then extracting band-power based features from these segmented signals. A classification algorithm, specifically the k-nearest neighbor algorithm, was then used to classify the features. The result was a remarkable classification accuracy (CA) of 95.54%±1.31 when using the active brain region within the hybrid model. These results underline the effectiveness of the proposed approach, as it outperformed both standalone EEG and NIRS modalities in terms of CA by 5.19% and 40.90%, respectively. Furthermore, the results confirm the considerable potential of the method in classifying EEG + NIRS signals recorded during tasks such as reading text while scrolling in different directions, including right, left, up and down. This research heralds a promising step towards enhancing the capabilities of BCI systems by harnessing the synergistic power of EEG and NIRS technologies.