A simple and efficient automated diagnostic model for Schizophrenia detection based on Hjorth Complexity

Abstract

Schizophrenia is a type of psychiatric disorder characterized by distortion of thoughts, speech and actions. Automatic detection of schizophrenia via Electroencephalography (EEG) has been a widely explored area of research. This work is based on developing a simple as well as an accurate model for automatic diagnosis of schizophrenia by analyzing the EEG patterns. Multivariate EEG data, collected by M.V. Lomonosov Moscow State University, is utilized. A sliding window of 0.5 s duration with 50% overlap is applied to extract Hjorth parameters (Activity, Mobility, Complexity) from EEG of both the normal controls and those with schizophrenic symptoms. In addition, feature dimension reduction is applied to further optimize the model performance using minimum redundancy maximum relevance algorithm. This model performed better than the current state-of-the-art techniques by achieving an overall accuracy of 98.82% using Quadratic Discriminant Analysis (QDA) classifier.