Motor imagery EEG signal classification using novel deep learning algorithm.

Area of Science:

• Biomedical Engineering
• Neuroscience
• Machine Learning

Background:

• Electroencephalography (EEG) signal classification is vital for neurological disorder detection and cognitive state monitoring.
• Existing methods face challenges including signal noise, inter-subject variability, and real-time processing demands, leading to reduced performance.

Purpose of the Study:

• To propose a novel model for motor imagery (MI) EEG signal classification that overcomes current limitations.
• To enhance classification accuracy and adaptability in diverse EEG signal analysis conditions.

Main Methods:

• A hybrid preprocessing approach combining empirical mode decomposition (EMD) and continuous wavelet transform (CWT) for signal mode extraction and multi-resolution analysis.
• Spatial feature enhancement using source power coherence (SPoC) integrated with common spatial patterns (CSP).
• Classification using an adaptive deep belief network (ADBN) optimized with the Far and near optimization (FNO) algorithm.

Main Results:

• Achieved 95.7% accuracy, 96.2% recall, 95.9% precision, and 97.5% specificity on the BCI competition IV Dataset 2a.
• Delivered 94.1% accuracy, 94.0% recall, 93.6% precision, and 95.0% specificity on the Physionet dataset.
• Demonstrated superior performance compared to Convolutional Neural Networks (CNN), Long Short-Term Memory (LSTM), and Bidirectional LSTM (BiLSTM) algorithms.

Conclusions:

• The proposed model offers superior classification accuracy and adaptability for EEG signal analysis.
• The novel combination of EMD, CWT, SPoC, CSP, and ADBN provides a robust solution for motor imagery EEG classification.
• This approach holds significant potential for advancing applications in brain-computer interfaces and neurological monitoring.