EEG/fNIRS Based Workload Classification Using Functional Brain Connectivity and Machine Learning

Area of Science:

• Neuroscience and Cognitive Science
• Biomedical Engineering
• Machine Learning

Background:

• Accurate estimation of human mental workload is crucial for enhancing productivity and preventing accidents.
• Existing methods often rely on single physiological sensing modalities and univariate analysis of electroencephalography (EEG) data.
• There is a need for advanced techniques that integrate multiple data sources and employ sophisticated analytical approaches.

Purpose of the Study:

• To propose a novel framework for multi-level mental workload classification using hybrid EEG-functional near-infrared spectroscopy (fNIRS) data.
• To investigate the efficacy of bivariate functional brain connectivity (FBC) features in the time and frequency domains for EEG analysis.
• To leverage machine learning for improved mental workload estimation.

Main Methods:

• Utilized a hybrid approach combining EEG and fNIRS physiological sensing modalities.
• Employed bivariate functional brain connectivity (FBC) features in delta, theta, and alpha frequency bands for EEG analysis, moving beyond univariate power spectral density (PSD).
• Integrated fNIRS oxyhemoglobin (HbO) and deoxyhemoglobin (HbR) indicators and applied machine learning algorithms for classification.

Main Results:

• The hybrid EEG-fNIRS framework achieved significant improvements in mental workload classification accuracy: 77% for 0-back vs. 2-back and 83% for 0-back vs. 3-back on a public dataset.
• Topographic and heat-map visualizations revealed distinct brain regions contributing to workload discrimination between EEG and fNIRS.
• Identified optimal regions for discrimination: posterior midline occipital (POz) for alpha-band EEG and the right frontal region (AF8) for fNIRS.

Conclusions:

• The proposed hybrid EEG-fNIRS framework with bivariate FBC features offers a robust approach for estimating multi-level mental workload.
• The study highlights the complementary nature of EEG and fNIRS, with different brain regions showing superiority for each modality in workload assessment.
• This research provides a foundation for developing more accurate and reliable mental workload monitoring systems.