Computerized Multidomain EEG Classification System: A New Paradigm.

Area of Science:

• Neuroscience and Computational Science
• Signal Processing and Machine Learning

Background:

• Current electroencephalogram (EEG) signal classification methods are often domain-specific, limiting their adaptability across different research areas.
• A unified framework is needed to improve the cross-discipline applicability of EEG classification algorithms.

Purpose of the Study:

• To introduce a computer-aided broad learning EEG system (CABLES) for classifying six distinct EEG domains within a single sequential framework.
• To enhance the multi-role adaptability and classification accuracy of EEG analysis systems.

Main Methods:

• Development of CABLES, incorporating three novel modules: complex variational mode decomposition (CVMD), ensemble optimization-based feature selection (EOFS), and t-distributed stochastic neighbor embedding-based samples reduction (tSNE-SR).
• Extensive experiments conducted on seven diverse EEG datasets using various neural network, extreme learning machine, and machine learning classifiers.
• A 10-fold cross-validation strategy was employed to evaluate the system's performance.

Main Results:

• CABLES achieved high classification accuracies across multiple datasets: 99.1% (motor imagery A), 97.8% (motor imagery B), 94.3% (slow cortical potentials), 91.5% (epilepsy), 98.9% (alcoholic EEG), 95.3% (schizophrenia EEG), and 92% (another dataset).
• The proposed framework demonstrated superior performance compared to existing domain-specific methods.

Conclusions:

• The CABLES framework offers superior classification accuracy and multi-role adaptability compared to existing domain-specific EEG classification methods.
• CABLES can be effectively utilized as an automated neural rehabilitation system, demonstrating its broad applicability and efficiency.