Neural Decoding of EEG Signals with Machine Learning: A Systematic Review
Maham Saeidi1, Waldemar Karwowski1, Farzad V Farahani1,2
1Computational Neuroergonomics Laboratory, Department of Industrial Engineering and Management Systems, University of Central Florida, Orlando, FL 32816, USA.
Brain Sciences
|November 27, 2021
Summary
This review explores artificial intelligence in electroencephalography (EEG) analysis. Machine learning and deep learning models show promise for decoding brain signals in tasks like motor imagery.
Area of Science:
- Neuroscience
- Computer Science
- Biomedical Engineering
Background:
- Electroencephalography (EEG) records brain electrical activity non-invasively.
- Artificial intelligence (AI), including machine learning (ML) and deep learning (DL), is increasingly used for EEG data analysis.
- Applications include pattern analysis, classification, and brain-computer interfaces.
Purpose of the Study:
- To systematically review recent advances in supervised ML and DL models for EEG signal decoding and classification.
- To provide a comprehensive overview of state-of-the-art EEG signal preprocessing and feature extraction techniques.
- To identify effective feature extraction methods and classifier recommendations.
Main Methods:
- Systematic literature search of academic databases from 2000 to present.
- Focus on supervised machine learning and deep learning models for EEG analysis.
- Analysis of preprocessing and feature extraction techniques.
Main Results:
- ML and DL applications in mental workload and motor imagery tasks are prominent.
- Convolutional neural networks (CNNs) dominate DL studies (75%).
- Support vector machines (SVMs) are frequently used in ML studies (36% accuracy).
- Wavelet transform is the most common feature extraction method across tasks.
Conclusions:
- ML and DL significantly advance EEG signal processing and interpretation.
- CNNs and SVMs are leading models for EEG classification.
- Wavelet transform is a versatile feature extraction technique for EEG data.
