Investigate Schizophrenia Classification Based on EEG Electrode Reduction Using Machine Learning Techniques

Schizophrenia is a mental disorder condition that causes patients to become distracted from reality.Over time, the patient loses his cognitive and social abilities to communicate with the outside world.Due to machine learning's strong ability to analyze complicated brain data, it has become an increasingly important tool in recent years.This study considers the brain's neurologic signals in the resting state in two scenarios to classify schizophrenia disease by electroencephalography (EEG).The performed scenarios were to investigate the impact of selecting electrodes randomly (5 electrodes and 8 electrodes) and comparing it with applying the principal component analysis (PCA), utilizing four algorithms to extract features: Fast Fourier Transform (FFT), Approximate Entropy (ApEn), Log Energy Entropy (LogEn), and Shannon Entropy (ShnEn).We used publicly available datasets with 19 EEG channels consisting of two classes, which are schizophrenia and health control, using a onesecond epoch window size.We applied a band-pass filter to decompose the EEG signals into five sub-bands.Also, the L2-normalization method has been applied to the derived features, which positively impacted the outcomes.The features were applied to three classifiers named K-nearest neighbor (KNN), support vector machine (SVM), and quadratic discriminant analysis (QDA).From all the scenarios, the five-electrode with random selection showed remarkable results of 99% using the SVM classifier in all evaluation metrics with LogEn+ Bandpass features.