A novel approach to schizophrenia Detection: Optimized preprocessing and deep learning analysis of multichannel EEG data

Abstract

Schizophrenia diagnosis, characterized by cognitive deficits, hallucinations, and delusions, poses challenges due to its complex nature. Electroencephalogram (EEG) signals provide insights into underlying brain states, enabling efficient and accurate diagnosis. This study introduces a novel approach, utilizing the Mutation-boosted Archimedes Optimization (MAO) algorithm to optimize preprocessing steps for improved EEG data quality. The spatial and temporal patterns are extracted from multichannel EEG data through Convolutional Neural Networks (CNNs) based Long Short-Term Memory (LSTM) network. Entropy-related features, including spectral and permutation entropy, are integrated to capture EEG complexity. The Schizophrenia Detection layer leverages these features for classification. The fitness function combines classification accuracy and noise reduction, optimizing preprocessing and classification performance. Experimental evaluation on a diverse dataset validates the effectiveness of the proposed approach. Achieving accuracy of 98.2%, F1-score of 97.2%, specificity of 97.5%, precision of 98.5%, and sensitivity of 98.9%, the method surpasses existing techniques. Moreover, the MAO algorithm enhances signal quality, evident through improved Signal-to-Noise Ratio (SNR), Signal-to-Interference Ratio (SIR), and reduced artifact rejection rates. Overall, this research demonstrates the potential of the Mutation-boosted Archimedes Optimization algorithm in enhancing EEG data preprocessing for accurate schizophrenia detection. The integration of deep learning architectures and advanced optimization techniques presents a promising avenue for improving mental disorder diagnoses through EEG signal analysis.