Electroencephalography-based diagnosis of schizophrenia using machine learning.

Area of Science:

• Neuroscience
• Computational Psychiatry
• Biomarker Discovery

Background:

• Schizophrenia presents a significant societal challenge, necessitating improved diagnostic methods.
• Quantitative electroencephalography (EEG) analysis offers potential for identifying objective biomarkers.
• Existing machine learning (ML) approaches for EEG-based schizophrenia detection lack optimized feature selection.

Purpose of the Study:

• To develop an automated routine for selecting optimal EEG biomarkers for schizophrenia diagnosis.
• To leverage machine learning parameter optimization for enhanced diagnostic accuracy.

Main Methods:

• Utilized the Waikato Environment for Knowledge Analysis (WEKA) software for automated ML parameter selection.
• Employed WEKA's "Supervised Attribute Selection" tool to identify significant EEG features.
• Analyzed EEG data to pinpoint specific frequency bands and brain regions relevant to schizophrenia.

Main Results:

• Achieved a high diagnostic accuracy of 93% in identifying schizophrenia patients.
• Identified specific EEG signal attributes, notably alpha and gamma frequencies.
• Highlighted the importance of frontal right, central, parietal, and occipital brain areas.

Conclusions:

• The proposed automated ML strategy effectively identifies schizophrenia patients with high accuracy.
• This approach can serve as a valuable ML tool to support clinical diagnosis.
• The identified biomarkers may offer insights into the neurophysiological mechanisms of schizophrenia.