Machine Learning in Electroconvulsive Therapy: A Systematic Review.

Area of Science:

• Neuroscience
• Computational Psychiatry
• Medical Informatics

Background:

• Predicting patient response to electroconvulsive therapy (ECT) remains a significant clinical challenge.
• Traditional statistical methods have limitations in leveraging complex, high-dimensional datasets.
• Machine learning (ML) offers a promising avenue for improving predictive accuracy in ECT.

Approach:

• A systematic review identified 26 articles applying ML to predict ECT response.
• Data sources included structural/functional imaging, clinical data, electroencephalography (EEG), and social media.
• Studies primarily focused on predicting treatment response in depression and psychosis.

Key Points:

• Brain imaging, particularly changes in the limbic system, shows predictive value for ECT response.
• Clinical factors like illness duration, severity, psychotic features, and cortisol levels predict response in depression.
• EEG-derived measures, such as transfer entropy, can predict relapse risk after ECT for psychosis.

Conclusions:

• ML approaches demonstrate potential for predicting ECT treatment outcomes.
• Integrating diverse data types (imaging, clinical, EEG) enhances predictive power.
• A collaborative, transdisciplinary approach is needed to translate these findings into clinical practice.