Related Experiment Video
Updated: Jun 12, 2025
06:40
Microstate and Omega Complexity Analyses of the Resting-state Electroencephalography
Published on: June 15, 2018
10.1K
Schizophrenia diagnosis based on diverse epoch size resting-state EEG using machine learning
Athar Alazzawı1, Saif Aljumaili1, Adil Deniz Duru2
1Electrical and Computer Engineering, School of Engineering and Natural Sciences, Altinbaş University, Istanbul, Turkey.
Peerj. Computer Science
|September 24, 2024
Summary
Early detection of schizophrenia is vital. This study introduces an electroencephalography (EEG) method using specific features and a support vector machine (SVM) classifier for accurate schizophrenia diagnosis.
Area of Science:
- Neuroscience and Psychiatry
- Biomedical Signal Processing
- Machine Learning in Healthcare
Background:
- Schizophrenia is a severe mental disorder impacting cognitive, social, and emotional functions.
- Early and accurate diagnosis of schizophrenia is critical for effective patient management and treatment.
- Electroencephalography (EEG) offers a non-invasive method for capturing brain activity.
Purpose of the Study:
- To develop and evaluate a novel method for classifying schizophrenia using resting-state EEG signals.
- To investigate the impact of data augmentation, feature reduction, and different window sizes on classification accuracy.
- To identify optimal feature extraction and classification algorithms for schizophrenia detection.
Main Methods:
- Collected resting-state EEG data from 28 subjects (14 schizophrenia, 14 healthy controls) using 19 channels at 250 Hz.
- Decomposed EEG signals into five sub-bands using band-pass filtering to enhance signal clarity and reduce artifacts.
- Employed feature extraction algorithms (FFT, ApEn, LogEn, ShnEn, kurtosis) with L2-normalization, tested with data augmentation (noise, stretching) and Minimum Redundancy Maximum Relevance (MRMR) feature reduction across 1, 2, and 5-second windows. Classification performed using KNN, SVM, QDA, and EC.
Main Results:
- The support vector machine (SVM) classifier achieved remarkable results with Log Energy entropy (LogEn) features using a 1-second window size.
- Data augmentation and MRMR feature reduction, combined with specific feature sets and window sizes, significantly influenced classification performance.
- The proposed method demonstrated notable improvements compared to recently published studies on similar datasets.
Conclusions:
- Accurate diagnosis of schizophrenia can be achieved through careful selection of appropriate EEG features and classification models.
- The combination of LogEn features, a 1-second window, and SVM provides a promising approach for automated schizophrenia detection.
- This study highlights the potential of advanced signal processing and machine learning techniques in psychiatric diagnostics.
