Decoding spatial attention by using cortical currents estimated from electroencephalography with near-infrared...

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Non-invasive brain-machine interfaces (BMIs) are crucial for real-world applications.
• Electroencephalography (EEG) and near-infrared spectroscopy (NIRS) are primary non-invasive methods.
• Current EEG and NIRS methods face challenges like signal mixing and time delays, limiting practical BMI utility.

Purpose of the Study:

• To develop a novel methodology for improving real-environment EEG-NIRS-based BMIs.
• To decode subjects' mental states using cortical currents estimated from EEG, enhanced by NIRS information.
• To overcome limitations of existing EEG and NIRS techniques for practical BMI applications.

Main Methods:

• Proposed a Variational Bayesian Multimodal EncephaloGraphy (VBMEG) methodology.
• Incorporated NIRS-based priors to capture event-related desynchronization from isolated cortical sources.
• Applied Bayesian logistic regression to decode mental states from sparsified current sources.

Main Results:

• The proposed EEG-NIRS decoder showed significant performance improvement compared to EEG-only methods.
• Decoding accuracy was enhanced by utilizing sparsely isolated current sources on the cortex.
• Neuroscientific validation confirmed the involvement of the intraparietal sulcus in spatial attention tasks.

Conclusions:

• The developed methodology offers a practical advancement for EEG-NIRS-based BMI applications.
• This approach holds potential for investigating brain activity in naturalistic, non-laboratory settings.
• The method effectively decodes mental states by integrating multimodal brain data, overcoming individual modality limitations.