Related Experiment Video
Updated: Jul 17, 2025
08:51
Statistical Modelling of Cortical Connectivity Using Non-invasive Electroencephalograms
Published on: November 1, 2019
5.7K
Analysis of functional connectivity using machine learning and deep learning in different data modalities from
Caroline L Alves1,2, Thaise G L de O Toutain3, Joel Augusto Moura Porto4
1University of São Paulo (USP), Institute of Mathematical and Computer Sciences (ICMC), São Paulo, Brazil.
Journal of Neural Engineering
|September 6, 2023
Summary
Machine learning models accurately diagnosed schizophrenia using brain imaging data. Electroencephalogram measures proved more effective than network analysis in detecting SCZ-related brain changes.
Area of Science:
- Neuroscience
- Computational Psychiatry
- Medical Imaging
Background:
- Schizophrenia (SCZ) affects 26 million globally, characterized by psychosis, hallucinations, and delusions.
- Machine learning (ML) and deep learning are increasingly used to automate SCZ diagnosis and study brain network dynamics.
Purpose of the Study:
- To develop a rigorous ML/deep learning approach for automated SCZ diagnosis using brain connectivity.
- To comprehend the topology and dynamics of brain networks in SCZ individuals.
- To integrate EEG measures with complex network analysis for novel insights.
Main Methods:
- Utilized functional magnetic resonance imaging (fMRI) and electroencephalogram (EEG) datasets.
- Combined EEG measures (Hjorth mobility, complexity) with complex network metrics.
- Applied ML and deep learning for evaluating connectivity matrices and network measures.
Main Results:
- Identified significant correlations in brain regions for SCZ patients compared to controls.
- The network diameter emerged as a crucial biomarker across modalities.
- SCZ brain networks showed reduced segregation and information distribution.
- EEG measures demonstrated superior performance over complex networks in capturing SCZ brain alterations.
Conclusions:
- The developed model achieved high accuracy (98.5% fMRI, 95.4% EEG) and AUC (100% fMRI, 95% EEG) for SCZ classification.
- Investigated specific brain connections and network measures contributing to classification accuracy.
- The study provides a robust framework for automated SCZ diagnosis and understanding brain network changes.
