Automatic classification of schizophrenia patients using resting-state EEG signals

Area of Science:

• Neuroscience
• Computational Psychiatry
• Signal Processing

Background:

• Schizophrenia is a severe mental disorder impacting multiple life aspects.
• Accurate and early diagnosis of schizophrenia is crucial for effective treatment.
• Electroencephalogram (EEG) signals offer a non-invasive window into brain activity.

Purpose of the Study:

• To develop and evaluate a new classification method for schizophrenia detection using EEG signals.
• To compare the performance of the proposed Adaptive Neuro-Fuzzy Inference System (ANFIS) with other machine learning models.
• To identify key EEG channels and frequency bands with high discriminatory power for schizophrenia.

Main Methods:

• Utilized EEG data from 14 schizophrenia patients and 14 healthy controls.
• Selected 16 EEG channels and applied a Butterworth filter to remove artifacts.
• Extracted four entropy-based and autoregressive features across five frequency sub-bands (Delta, Theta, Alpha, Beta, Gamma).
• Employed ANFIS, Support Vector Machine (SVM), and Artificial Neural Network (ANN) classifiers.

Main Results:

• The ANFIS classifier achieved near 100% accuracy in distinguishing schizophrenia patients from controls.
• SVM and ANN classifiers showed high accuracies of 98.89% and 95.59%, respectively.
• Specific EEG channels (e.g., alpha at O1, theta/delta at Fz/F8, gamma at Fp1) were identified as highly informative.

Conclusions:

• The proposed ANFIS-based method demonstrates exceptional accuracy for classifying schizophrenia from EEG signals.
• This approach can serve as a valuable decision support system (DSS) for identifying individuals with schizophrenia.
• The findings highlight the potential of advanced signal processing and machine learning in psychiatric diagnostics.