Classification of First-Episode Psychosis with EEG Signals: ciSSA and Machine Learning Approach.

Area of Science:

• Neuroscience
• Psychiatry
• Signal Processing

Background:

• First-episode psychosis (FEP) signifies the onset of severe psychiatric disorders, making early diagnosis crucial for effective intervention and better patient outcomes.
• Accurate and timely diagnosis of FEP presents a significant challenge in mental healthcare.
• Electroencephalography (EEG) signals offer a potential avenue for objective diagnostic markers in psychiatric conditions.

Purpose of the Study:

• To develop and validate a novel, high-performance classification method for the early diagnosis of FEP.
• To investigate the utility of circulant spectrum analysis (ciSSA) sub-band signals derived from EEG for FEP classification.
• To identify and leverage significant features from EEG signals for machine learning-based FEP detection.

Main Methods:

• EEG signals were analyzed using circulant spectrum analysis (ciSSA) to extract sub-band features.
• The LASSO method was employed for feature selection, focusing on entropy, frequency, and statistical characteristics from ciSSA sub-bands and their combinations.
• Machine learning models, including ensemble methods, Support Vector Machine (SVM), and Artificial Neural Network (ANN), were utilized for classification.

Main Results:

• The hybrid features from EEG ciSSA sub-bands combined with the SVM classifier yielded exceptional performance.
• Key performance metrics included an Area Under the Curve (AUC) of 0.9893, 96.23% accuracy, 0.966 sensitivity, 0.956 specificity, 0.9667 precision, and 0.9666 F1 score.
• The findings underscore the effectiveness of the ciSSA-based approach for classifying FEP using EEG data.

Conclusions:

• The ciSSA-based method demonstrates significant potential as an effective tool for the early and accurate classification of FEP from EEG signals.
• This approach offers a promising advancement in objective diagnostic strategies for first-episode psychosis.
• Further research can explore the clinical integration of this EEG analysis technique for improved mental health diagnostics.