A pilot study to determine whether machine learning methodologies using pre-treatment electroencephalography can...

Area of Science:

• Neuroscience
• Psychiatry
• Machine Learning

Background:

• Schizophrenia is a chronic mental disorder often requiring long-term treatment.
• Clozapine is an effective treatment for treatment-resistant schizophrenia, but predicting response is challenging.
• Electroencephalography (EEG) measures brain activity and may contain predictive biomarkers.

Purpose of the Study:

• To determine if machine learning (ML) applied to pre-treatment electroencephalography (EEG) data can predict clozapine response in chronic schizophrenia.
• To identify relevant EEG features for predicting treatment outcomes.

Main Methods:

• Collected pre-treatment EEG data from 37 adults with schizophrenia.
• Employed a feature selection scheme to identify statistically relevant EEG features.
• Utilized kernel partial least squares regression for response prediction, validated with the Positive and Negative Syndrome Scale (PANSS).

Main Results:

• Identified a set of discriminating EEG features capable of predicting clozapine response.
• Demonstrated significant clustering of EEG features between clozapine responders and non-responders.
• Achieved a minimum prediction accuracy of 85% using leave-one-out cross-validation and independent sample testing.

Conclusions:

• Pre-treatment EEG analysis, using advanced ML, can predict clinical response to clozapine in treatment-resistant schizophrenia.
• This novel EEG analysis approach shows promise in assisting clinicians with treatment efficacy determination.
• Replication in larger populations is warranted to confirm the clinical utility of this predictive method.