Related Experiment Video
Updated: Apr 24, 2026
09:47
Author Spotlight: Advancing Alzheimer's Research – Exploring Early Detection and Multi-Omics Approaches
Published on: December 15, 2023
2.0K
A comparative evaluation of EEG-based deep learning models for schizophrenia detection with cross-dataset validation
Jagan S J1, Smrithy G S1, Balaji Chandrasekaran2
1School of Computer Science and Engineering, Vellore Institute of Technology, Chennai, India.
Neurological Research
|April 22, 2026
Summary
Deep learning models combined with electroencephalography (EEG) spectral analysis show promise for automated schizophrenia detection. While models performed similarly, spectral features in CNN architectures are particularly beneficial for diagnosis.
Area of Science:
- Neuroscience
- Artificial Intelligence
- Medical Diagnostics
Background:
- Schizophrenia is a neuropsychiatric disorder impacting brain function, detectable via electroencephalography (EEG).
- Automating schizophrenia detection using EEG analysis is crucial for timely diagnosis and treatment.
Purpose of the Study:
- To comparatively evaluate deep learning models for automated schizophrenia detection using EEG time-frequency and spectral analysis.
- To assess the effectiveness of various deep learning architectures and feature extraction methods.
Main Methods:
- Merged two EEG datasets (934 samples, 237 subjects) and applied Independent Component Analysis (ICA).
- Derived time-frequency representations using Continuous Wavelet Transform (CWT) and Fast Fourier Transform (FFT).
- Evaluated six deep learning models (CNN variants, ResNet, Transformer) with data augmentation and class balancing.
Main Results:
- Variations in model performance were observed but were not statistically significant.
- Deep learning models utilizing spectral features, particularly within CNN architectures, showed promise for schizophrenia detection.
Conclusions:
- Combining time-frequency analysis with deep learning, especially spectral features in CNNs, benefits EEG-based schizophrenia diagnosis.
- Model interpretability is key for clinical translation; future work includes multimodal neuroimaging and explainability frameworks.