Machine Learning Approach to Diagnose Schizophrenia Based on Effective Connectivity of Resting EEG Data

Schizophrenia is a severe mental disorder associated with neurobiological deficits. Despite the fact that the brain activity during tasks (i.e. P300 activities) are considered as biomarkers for diagnosing schizophrenia, brain activities at rest has the potential to reveal an intrinsic dysfunctionality in patients with schizophrenia and can be used to understand the cognitive deficits in these patients. In this study, we develop a machine learning (ML) algorithm based on eye closed resting-EEG data sets aiming to distinguish 63 schizophrenic patients (SCZs) from 70 healthy controls (HCs). The ML algorithm has three steps. In the first step an effective connectivity named symbolic transfer entropy (STE) is applied to the EEG waveforms. In the second step brain network properties are constructed from STE. In the third step the ML algorithm is applied to brain network properties to determine whether a set of features can be found that successfully discriminates SCZ from HC. The findings of this study revealed that the most discriminating features are shortest-path-length between different brain regions that could achieve an accuracy of 96.15%, with a sensitivity of 100% and specificity of 92.86% using 20% of data samples as test dataset that is not used for training. These findings imply that resting EEG could contribute to our ability to distinguish SCZs from HCs, and that the STE effective connectivity may prove to be a promising tool for the clinical diagnosis of schizophrenia.