EEG workload estimation and classification: a systematic review.

Area of Science:

• Neuroscience and Artificial Intelligence
• Human-Computer Interaction
• Cognitive Science

Background:

• Electroencephalography (EEG) is crucial for assessing cognitive workload in diverse fields.
• Machine learning (ML) and deep learning (DL) are increasingly used to build accurate EEG-based workload models.
• A systematic review is needed to consolidate research on EEG workload estimation using ML/DL.

Purpose of the Study:

• To systematically review and compile research on cognitive workload estimation and classification using EEG data with ML and DL techniques.
• To identify common ML/DL algorithms, study designs, and performance metrics in this field.

Main Methods:

• Systematic literature search conducted across major scientific databases (SpringerLink, ACM, IEEE, PubMed, Science Direct) up to February 16, 2024.
• Studies selected based on predefined inclusion criteria following PRISMA guidelines.
• Data extraction focused on study design, participant demographics, EEG features, ML/DL algorithms, and performance metrics.

Main Results:

• 33 out of 125 initially identified papers were included in the final analysis.
• Support Vector Machines (SVM), Convolutional Neural Networks (CNN), and Recurrent Neural Networks (RNN) were the most frequently used ML/DL techniques.
• Higher sampling frequencies in EEG data generally correlated with improved model accuracy; SVM, CNN, and hybrid networks showed robustness.

Conclusions:

• ML effectively estimates mental workload from EEG data across various applications.
• Future advancements require multimodal data integration, standardization, and real-world validation.
• Addressing ethical considerations and exploring novel EEG properties will enhance ML/DL models for improved human-computer interaction and performance assessment.