Electroencephalograph (EEG) signal analysis for the Detection of Schizophrenia using Empirical Wavelet Transform

Schizophrenia (SCZ) is a severe mental disorder that affects behavior, speech, mood etc. of people across the world. Early detection of SCZ can play a vital role in planning the treatment for patients. Recent studies confirms that Electroencephalography (EEG) signal can be used effectively for detection of SCZ. This work attempts to propose a simple machine learning based model with improved performance for detection of SCZ. The study was conducted on 19 channel rest state EEG signal recording of total 16 subjects out of which 8 were SCZ and 8 healthy controls (HC). After acquiring the signal, preprocessing is done and signal is decomposed using Empirical Wavelet Transform (EWT) to analyze the EEG components. 3 different entropy features were calculated over the decomposed signal. The features of selected significant mode function were applied to the classifiers named as support vector machine (SVM), k-nearest neighbor (KNN), linear discriminant (LD) and neural network. Results indicates that EWT could be a useful method for analysis of EEG signal and classification problems as various classifiers namely Fine KNN, Quadratic SVM and Wide Neural Network achieved the best classification accuracy of 87.5% with 5-fold data division protocol.