Related Experiment Video
Updated: Jul 2, 2025
Mapping Cortical Dynamics Using Simultaneous MEG/EEG and Anatomically-constrained Minimum-norm Estimates: an Auditory Attention Example
Published on: October 24, 2012
Brain network analysis of working memory in schizophrenia based on multi graph attention network
Ping Lin1, Geng Zhu2, Xinyi Xu1
1College of Medical Instruments, Shanghai University of Medicine & Health Sciences, Shanghai 201318, China; College of Health Science and Engineering, University of Shanghai for Science and Technology, Shanghai 200093, China.
Schizophrenia (SZ) patients exhibit altered brain activity during working memory tasks, including reduced P3 wave amplitude and distinct neural oscillations. A novel graph attention network model accurately classifies SZ based on these brain connectivity differences.
Area of Science:
- Neuroscience
- Psychiatry
- Computational Neuroscience
Background:
- Cognitive impairment, particularly in working memory, is a core feature of schizophrenia (SZ).
- Understanding the neural underpinnings of these deficits during cognitive tasks is crucial for diagnosis and treatment.
- Event-related potentials (ERPs) and brain functional connectivity offer insights into neural processing.
Purpose of the Study:
- To investigate brain changes in schizophrenia patients during a working memory task using ERP and functional connectivity analysis.
- To develop and validate a machine learning model for classifying schizophrenia based on neurophysiological data.
Main Methods:
- Time-domain and time-frequency analysis of event-related potentials (ERPs) during a 0-back task in SZ patients and healthy controls (HC).
- Phase lag index (PLI) was used to assess brain functional connectivity between different brain regions.
- A multi graph attention network model with adaptive initial residual (AIR) was proposed for SZ classification.
Main Results:
- SZ patients showed significantly lower P3 wave amplitude compared to HC (p < 0.05).
- Distinct patterns of neural oscillations (θ, α, β bands) were observed in SZ and HC groups post-stimulation.
- Significant differences in functional connectivity between parietal and frontotemporal lobes were identified between SZ and HC groups (p < 0.05).
- The proposed multi graph attention network achieved high classification accuracy (90.90% on an open dataset, 78.57% on the 0-back dataset).
Conclusions:
- Schizophrenia is associated with specific alterations in event-related potentials and brain functional connectivity during working memory tasks.
- The developed graph attention network model demonstrates significant potential for objective classification of schizophrenia based on neurophysiological markers.
- These findings contribute to a better understanding of the neural basis of cognitive deficits in schizophrenia.
