Related Experiment Video
Updated: Jul 17, 2026
12:00
Investigating the Effects of Antipsychotics and Schizotypy on the N400 Using Event-Related Potentials and Semantic Categorization
Published on: November 19, 2014
Transformer architecture for diagnosing schizophrenia disabilities through EEG analysis
Nizar Alsharif1, Nadhem Ebrahim2, Abdullah H Al-Nefaie3
1Department of Computer Science, Al-baha University, Albaha, Saudi Arabia.
Frontiers in Physiology
|July 16, 2026
Summary
Deep learning models, specifically Transformer architectures, show promise in accurately detecting schizophrenia using electroencephalography (EEG) data. This automated approach could lead to earlier diagnosis and improved patient outcomes.
Area of Science:
- Neuroscience
- Artificial Intelligence
- Psychiatry
Background:
- Schizophrenia diagnosis presents significant clinical challenges.
- Electroencephalography (EEG) offers a potential neurophysiological marker for psychiatric disorders.
- Automated analysis of EEG data can aid in objective diagnostic tools.
Purpose of the Study:
- To develop and evaluate deep learning models for automated schizophrenia detection using EEG.
- To compare the performance of LSTM and Transformer-based architectures for this task.
- To identify key features from EEG signals discriminative of schizophrenia.
Main Methods:
- EEG data from schizophrenia patients and healthy controls were preprocessed, including filtering and artifact removal.
- Temporal and spectral features were extracted from EEG signals.
- Feature selection methods (ANOVA, Mutual Information, Random Forest) identified discriminative features.
- LSTM and Transformer deep learning models with attention mechanisms and temporal convolutional networks were developed and evaluated.
Main Results:
- A Transformer-based deep learning model achieved 98.20% accuracy, 98.11% sensitivity, and 98.12% specificity in detecting schizophrenia.
- An LSTM-based model showed strong performance with 95.65% accuracy.
- The Transformer model demonstrated superior performance in cross-validation, highlighting its potential for robust schizophrenia detection.
Conclusions:
- Transformer architectures show significant potential for the accurate and automated detection of schizophrenia via EEG analysis.
- This approach could provide clinicians with an objective tool for earlier and more precise schizophrenia diagnosis.
- Timely intervention based on objective diagnostic support may lead to improved treatment outcomes for schizophrenia patients.
