Microstate Graphs: A Node-Link Approach to Identify Patients with Schizophrenia

An accurate identification of schizophrenia is important for its early diagnosis and treatment. Due to the complexity and inherent heterogeneity of this disorder, an accurate marker has not been established yet. In this paper is proposed a new methodology for the detection of schizophrenia using EEG microstate analysis and graph theory. The proposed method, called Microstate Graphs, allows modeling and interpreting each microstate as a complex network, which permits to identify the effect of schizophrenia on some characteristics of the built networks. In an experiment carried out in a public dataset, the proposed method was useful to identify schizophrenic patients with an accuracy of (\(91.67 \pm 2.06 \))% using an MLP trained with metrics extracted from microstate networks: assortativity, small-worldness, and local efficiency, indicating that the proposed method is promising.