Schizophrenia Classification Using Resting State EEG Functional Connectivity: Source Level Outperforms Sensor Level.

Area of Science:

• Neuroscience
• Computational Psychiatry
• Biomedical Engineering

Background:

• Disrupted brain connectivity is a hallmark of schizophrenia.
• Existing EEG-based classification methods struggle with signal interference (volume conduction).
• Sensor-level EEG data mixes signals, creating inaccurate connectivity estimates.

Purpose of the Study:

• To enhance the accuracy of distinguishing schizophrenia patients from healthy individuals using EEG.
• To overcome the limitations of volume conduction in EEG functional connectivity analysis.
• To explore source-level EEG signal analysis for improved machine learning applications.

Main Methods:

• Transformed resting-state EEG time series from sensor to source level using source reconstruction.
• Calculated functional connectivity networks using phase lag index at both sensor and source levels.
• Employed complex network analysis and feature selection for classifier input, using logistic regression for classification across five frequency bands.

Main Results:

• Connectivity measures derived from source-level EEG data yielded better classifier performance than sensor-level data.
• The theta frequency band, combined with source-level connectivity, provided the most effective features for classification.
• Source transformation significantly improved the ability to differentiate schizophrenia patients from controls.

Conclusions:

• Transforming scalp EEG to source-level signals enhances functional connectivity analysis.
• Source-level EEG connectivity offers superior features for machine learning in psychiatric research.
• This approach holds promise for improving diagnostic tools for schizophrenia.