Abstract
Schizophrenia (SZ) is an illness of the mind. Electroencephalogram (EEG) is an excellent notion of forecasting abnormalities and cognitive tasks as part of the study of mental illness. This scholarly research aimed to develop an Artificial Intelligence System (AIS) to research and classify the EEG signal features of healthy and Schizophrenia patients. The developed AIS was utilized to assess a 19-channel EEG signal collected from normal and Schizophrenia patients. The EEG signal is decomposed using Multi-level Discrete Wavelet Transformation (MDWT) into distinct EEG rhythms, specifically theta, alpha, beta, and gamma. Eight statistical feature extraction was then implemented to calculate 2128 features of healthy and 2128 features of schizophrenia patients from the Theta-EEG rhythm. And finally, ten machine-learning state-of-the-art classifiers are stacked together to classify the 4256 statistical features with different kernels. The experimental outcomes; the performance of (a) the Ada-Boost Classifier with Central electrodes is 85.71%, (b) the Gradient-Boosting Classifier with Parietal-Occipital electrodes is 80.00%, (c) the Decision-Tree Classifier & Ada-Boost Classifier with Frontal-Prefrontal electrodes is 76.67%, and (d) the XGBoost Classifier with the combination of the Central, Parietal-Occipital and Frontal-Prefrontal electrodes is 78.75%. The research achievement; (A) “In 10–20 Electrodes system of the brain architecture, Central electrodes are the better choice as compared to the Parietal-Occipital, and Frontal-Prefrontal electrodes for the EEG-based investigation of SZ,” (B) “Parietal-Occipital electrodes are the better choice as compared to the Frontal-Prefrontal electrodes for the EEG-based investigation of SZ,” (C) “Theta-EEG rhythm is the most relevant EEG rhythm for the diagnosis of SZ.”
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have