A Survey of Methods and Architectures for EEG-based Classification of Neuropsychiatric Disorders Using Deep Sparse Neural Networks with Gooseneck Barnacle Optimization Algorithm

Abstract

Electroencephalography (EEG)-based classification of neuropsychiatric disorders has emerged as a critical research domain due to its non-invasive nature and ability to capture real-time brain activity. Recent advancements in deep learning, particularly deep sparse neural networks, have significantly improved classification accuracy by enabling automatic feature extraction and reducing redundancy in high-dimensional EEG data. This paper presents a comprehensive survey of methods and architectures employed for EEG-based diagnosis of neuropsychiatric disorders such as depression, schizophrenia, and Alzheimer’s disease. Furthermore, the integration of bio-inspired optimization algorithms, specifically the Gooseneck Barnacle Optimization Algorithm (GBOA), is explored to enhance model convergence, feature selection, and parameter tuning. The survey highlights key developments, including convolutional neural networks (CNN), recurrent neural networks (RNN), graph neural networks (GNN), and hybrid deep architectures. Studies indicate that deep learning models achieve average classification accuracies exceeding 90%, with CNN-based models being the most widely adopted . However, challenges such as data variability, limited datasets, and interpretability persist.

This work provides a comparative analysis of existing approaches, identifies research gaps, and discusses future directions for developing efficient, interpretable, and optimized EEG-based diagnostic systems using sparse architectures and evolutionary optimization techniques.