EEG power spectrum analysis for schizophrenia during mental activity.

Area of Science:

• Neuroscience
• Biomedical Engineering
• Psychiatry

Background:

• Cognitive dysfunction is a primary characteristic of schizophrenia.
• Brain structural and functional abnormalities, detectable via Electroencephalogram (EEG), underlie cognitive deficits.
• EEG analysis offers a non-invasive method to investigate brain activity in schizophrenia.

Purpose of the Study:

• To analyze Electroencephalogram (EEG) power spectral density in schizophrenia patients during mental activity.
• To identify specific EEG band power features for distinguishing schizophrenia subjects from healthy controls.
• To evaluate the efficacy of machine learning classifiers (BPN and SVM) in diagnosing schizophrenia based on EEG features.

Main Methods:

• EEG data recorded from 52 schizophrenia patients and 29 normal subjects under resting and mental activity conditions (two modified oddball paradigms).
• EEG signals filtered using FIR bandpass to extract delta, theta, alpha, and beta bands; absolute power calculated using Welch's power spectral density method.
• Statistical analysis (Student's t-test) identified significant differences (p < 0.05); features used for classification with Backpropagation Neural Network (BPN) and Support Vector Machine (SVM).

Main Results:

• Schizophrenia subjects exhibited lower absolute alpha band power and higher absolute beta band power compared to normal subjects during rest and mental tasks.
• Significant EEG features were identified across all recording conditions.
• The Support Vector Machine (SVM) classifier achieved a maximum sensitivity of 91% when combining features from all recording conditions.

Conclusions:

• Mental activity EEG analysis, specifically absolute band powers, provides valuable features for classifying schizophrenia.
• The findings support the use of EEG-based biomarkers for identifying schizophrenia.
• This study highlights the potential of non-invasive EEG analysis in schizophrenia diagnosis and research.