Advanced EEG signal classification for neural prosthetic devices using metaheuristic and deep learning techniques

Area of Science:

• Biomedical Engineering
• Computational Neuroscience
• Machine Learning

Background:

• High-dimensional electroencephalography (EEG) signal classification for neural prosthetics is hindered by redundant features, impacting classifier generalization and efficiency.
• Accurate decoding of EEG signals is crucial for reliable real-time operation of neural prosthetic control systems.

Purpose of the Study:

• To present a unified, optimal-driven framework to address feature redundancy and improve EEG-based motor imagery (MI) signal decoding.
• To enhance the accuracy and computational efficiency of EEG signal classification for advanced neural prosthetics.

Main Methods:

• A novel feature selection model integrating the Coati Optimization Algorithm (COA) with opposition-based learning was developed for dynamic, parameter-free adaptation in high-dimensional spaces.
• Optimized feature subsets were used to train various classifiers, including Support Vector Machines (SVM), Random Forests (RF), Convolutional Neural Networks (CNN), and Recurrent Neural Networks (RNN).
• The framework was validated on benchmark EEG datasets, specifically the PhysioNet Motor Movement/Imagery dataset.

Main Results:

• The COA + CNN model achieved the highest classification accuracy at 96.8%, with precision, recall, and F1-score exceeding 96%.
• This represents a significant 6.5% improvement over existing feature selection techniques.
• The proposed method substantially outperformed conventional algorithms like Particle Swarm Optimization (PSO) and Genetic Algorithms (GA), as well as filter-type methods (mRMR, ReliefF).

Conclusions:

• Combining metaheuristic feature selection (COA) with deep learning architectures (CNN) offers a powerful approach for accurate EEG signal classification.
• The COA-based method provides a robust, computationally efficient, and scalable solution for high-accuracy classification, vital for future neural prosthetics.
• This framework promotes enhanced reliability and real-time capabilities in neural prosthetic control systems.