Enhanced schizophrenia detection using multichannel EEG and CAOA-RST-based feature selection
Abstract Schizophrenia is a mental disorder characterized by hallucinations, delusions, disorganized thinking and behavior, and inappropriate affect. Early and accurate diagnosis of schizophrenia remains a challenge due to the disorder’s complex nature and the limitations of state-of-the-art techniq...
Saved in:
| Main Authors: | , , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Nature Portfolio
2025-07-01
|
| Series: | Scientific Reports |
| Subjects: | |
| Online Access: | https://doi.org/10.1038/s41598-025-05028-7 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Abstract Schizophrenia is a mental disorder characterized by hallucinations, delusions, disorganized thinking and behavior, and inappropriate affect. Early and accurate diagnosis of schizophrenia remains a challenge due to the disorder’s complex nature and the limitations of state-of-the-art techniques. It is evident from the literature that electroencephalogram (EEG) signals provide valuable insights into brain activity, but their high dimensionality and complexity pose remain key challenges. Thus, our research introduces a novel approach by integrating the multichannel EGG, Crossover-Boosted Archimedes Optimization Algorithm (CAOA), and Rough Set Theory (RST) for schizophrenia detection. It is a four-stage model. In the first stage, Raw EGG data is collected. The data is passed to the next stage, which is called data preprocessing. This is used for artifact removal, band-pass filtering, and data normalization. The preprocessed data passed to the next stage. In the feature extraction stage, feature selection is performed using CAOA. In addition, classification is performed using a Support Vector Machine (SVM) based on features extracted through Multivariate Empirical Mode Function (MEMF) and entropy measures. The data interpretation stage displays the results to the end user using the data interpretation stage. We experimented and tested our proposed model using real EEG datasets. The simulation results prove that the proposed model achieved an average accuracy of 94.9%, sensitivity of 93.9%, specificity of 96.4%, and precision of 92.7%. Thus, our proposed model demonstrates significant improvements over state-of-the-art methods. In addition, the integration of CAOA and RST effectively addresses the challenges of high-dimensional EEG data, helps optimize the feature selection process, and increases accuracy. In future work, we suggest incorporating large-size datasets that include more diverse patient groups and refining the model with advanced machine-learning models and techniques. |
|---|---|
| ISSN: | 2045-2322 |