A Comprehensive Review of EEG-Based Classification of Neuropsychiatric Disorders Using Deep Sparse Neural Networks with Gooseneck Barnacle Optimization Algorithm

Abstract

Neuropsychiatric disorders such as depression, schizophrenia, anxiety, and bipolar disorder pose significant challenges for early diagnosis due to their complex and heterogeneous nature. Electroencephalography (EEG) has emerged as a promising non-invasive tool for detecting abnormal brain activity associated with these disorders. Recent advancements in deep learning have enabled automated feature extraction from EEG signals, improving diagnostic accuracy and reducing reliance on handcrafted features. Deep sparse neural networks have gained attention for their ability to reduce model complexity while preserving essential features, making them suitable for high-dimensional EEG data. Sparse learning techniques eliminate redundant connections, enhancing computational efficiency and interpretability. Additionally, optimization algorithms such as the Gooseneck Barnacle Optimization (GBO) algorithm have been introduced to improve model training by optimizing weight parameters and avoiding local minima. This review provides a comprehensive analysis of EEG-based neuropsychiatric disorder classification methods, focusing on deep sparse neural networks and optimization strategies. Comparative analysis highlights the advantages of hybrid architectures integrating deep learning, sparsity, and optimization techniques. The paper also discusses key challenges such as data variability, noise, and interpretability, and outlines future research directions for developing robust and scalable diagnostic systems.