Integrating EEG and NIRS improves BCI performance during motor imagery

Abstract

Hybrid brain-signal could theoretically achieve a more accurate decoding in brain-computer interfaces (BCIs). This study aims to improve the BCI performance by integrating electroencephalographic and cerebral hemodynamic patterns. Sixteen volunteers participated in a seven-session motor imagery (MI) based visual-haptic neurofeedback training (NFT) experiment. Electroencephalogram (EEG) and near infrared spectroscopy (NIRS) signals were synchronously recorded during the transient NFT. Time-frequency and topology analysis demonstrated that repetitive and continuous visual-haptic NFT conditions could induce an enhancement on cortical activations. Additionally, a classifier calibration strategy was proposed by integrating both the EEG and NIRS information into a strong classifier. The results revealed that the classifier constructed on integrating EEG and NIRS patterns was significantly superior to that only with independent information, achieving ∼14% and 5% improvement respectively and reaching 84.5% in mean classification accuracy. Therefore, integrating EEG and NIRS patterns is an effective classifier calibration strategy to improve the MI-BCI performance, and could also be applied to other BCI paradigms.