Study of Network Information Interactions in Schizophrenia Magnetoencephalograms Based on Permutation Transfer Entropy

Academic Background

Schizophrenia (SCZ) is a mental disorder characterized by persistent delusions and hallucinations, disorganized thinking, and inconsistent behavior, often leading to significant impairment in the perception of reality. With the rapid development of modern neuroimaging techniques, extensive datasets support research on neurological and mental disorders. Magnetoencephalography (MEG), a neuroimaging technique known for its high spatial and temporal resolution, can capture the nonlinear characteristics of the brain’s electromagnetic signals, and thus is employed to explore the information interactions in schizophrenia (SCZ).

Source

This paper, entitled “Networked information interactions in schizophrenia magnetoencephalograms based on permutation transfer entropy,” is authored by Qiong Wang, Xinran Yang, Wei Yan, Jiafeng Yu, and Jun Wang, from Nanjing University of Posts and Telecommunications, Nanjing Brain Hospital, and various institutions at the University of Texas. The article has been accepted by the journal “Biomedical Signal Processing and Control” and is expected to be published on February 3, 2024.

Research Objectives and Significance

This study aims to explore the information interactions in the MEG data of schizophrenia (SCZ) patients and healthy controls (HCs) by constructing resting-state brain networks using a method based on permutation transfer entropy. The study particularly considers the impact of equal values in the permutation and probability distribution process and quantifies network characteristics such as weight, complexity, and non-equilibrium to characterize the schizophrenia MEG network, thereby extending the exploration of the pathological and physiological mechanisms of schizophrenia.

Research Process

Subjects and Data Collection

The experimental data were collected from 31 subjects at Nanjing Medical University’s Affiliated Brain Hospital, including 17 schizophrenia patients and 14 healthy controls. All subjects were right-handed and factors such as other mental disorders and serious brain injuries were excluded. MEG recordings were performed using a CTF MEG system at a sampling frequency of 1200 Hz for 2 minutes.

Data Processing and Preprocessing

To ensure accuracy, subjects’ ECG and EOG signals were monitored during the experiment. Data processing used the FieldTrip toolbox for filtering and processing, and the 275 MEG channels were divided into 14 regions based on MEG partition templates and physiological significance of brain areas.

Permutation Transfer Entropy Calculation

Permutation transfer entropy with equal value consideration was employed to quantify the information interactions between brain regions. For each MEG dataset, brain regions were defined as network nodes, and a representative time series was obtained by averaging the MEG time series of all channels within each brain region. The MEG networks under different scenarios were constructed with embedding dimension m and delay time 𝜏 set, and the significance of network connections was assessed using hypothesis testing.

Network Information Flow Analysis

The inward, outward, and total information flow of each brain region was calculated, and independent sample t-tests were used to detect group differences. It was found that the level of information interactions in the brain regions of schizophrenia patients was generally lower, with the primary differences concentrated in the frontal and parietal regions, consistent with many previous studies.

Complexity and Non-Equilibrium Analysis

Shannon entropy was used to calculate the complexity of information flow within brain regions. The results showed that the information flow complexity in the SCZ group was significantly higher than that of the HCs group. However, the non-equilibrium analysis