Related Experiment Video
Updated: Aug 8, 2025
13:18
Conducting Concurrent Electroencephalography and Functional Near-Infrared Spectroscopy Recordings with a Flanker Task
Published on: May 24, 2020
7.8K
Raw Electroencephalogram-Based Cognitive Workload Classification Using Directed and Nondirected Functional
Anmol Gupta1, Ronnie Daniel2, Akash Rao3
1Department of Computer Science and Engineering, Indian Institute of Technology Roorkee, Roorkee, India.
Big Data
|February 27, 2023
Summary
This study shows that combining functional connectivity algorithms with deep learning, specifically Phase Transfer Entropy (PTE) and BrainNetCNN, can accurately classify cognitive workload levels from electroencephalogram (EEG) data, achieving 99.50% accuracy.
Area of Science:
- Neuroscience
- Computer Science
- Biomedical Engineering
Background:
- Electroencephalogram (EEG)-based brain-computer interfaces (BCIs) offer potential for controlling devices and proactive health management.
- EEG signals are inherently noisy and variable, posing challenges for real-time data processing and robust cognitive state monitoring.
- Existing methods struggle to accurately reflect neuronal dynamics amidst high EEG variability, particularly for passive BCIs that track cognitive workload.
Purpose of the Study:
- To evaluate the efficacy of combining functional connectivity algorithms with deep learning for classifying cognitive workload levels.
- To compare the performance of Phase Transfer Entropy (PTE) and Mutual Information (MI) as functional connectivity measures.
- To assess the classification accuracy of the BrainNetCNN deep learning model using these connectivity measures.
Main Methods:
- Acquired 64-channel EEG data from 23 participants performing the n-back task at low (1-back), medium (2-back), and high (3-back) workload conditions.
- Extracted functional connectivity matrices using Phase Transfer Entropy (PTE) and Mutual Information (MI).
- Classified cognitive workload levels using the BrainNetCNN deep learning model.
Main Results:
- Mutual Information (MI) with BrainNetCNN achieved 92.81% classification accuracy.
- Phase Transfer Entropy (PTE) with BrainNetCNN achieved a significantly higher accuracy of 99.50%.
- PTE demonstrated superior robustness to data artifacts and ability to detect functional connectivity across lags.
Conclusions:
- The combination of PTE and BrainNetCNN offers a highly accurate and robust method for classifying cognitive workload from EEG data.
- This approach holds promise for developing advanced passive BCIs for real-time cognitive state monitoring.
- The findings highlight the potential of directed functional connectivity measures in improving BCI performance.
