An artificial intelligence-based platform for personalized predictions of Metacognitive Training effectiveness

Area of Science:

• Psychiatry
• Computer Science
• Artificial Intelligence

Background:

• Metacognitive Training (MCT) is a therapeutic approach for psychosis.
• Personalizing treatment plans is crucial for improving patient outcomes.
• Existing decision support systems may lack comprehensive predictive capabilities.

Purpose of the Study:

• To develop and evaluate a machine learning (ML)-based platform for predicting MCT effectiveness.
• To create a prototype decision support system for clinicians treating psychosis patients.
• To enhance treatment personalization through data-driven insights.

Main Methods:

• Integration of eight ML models to predict MCT effectiveness.
• Utilizing a wide range of mental health questionnaires to assess diverse psychological symptoms.
• Implementation of explainable AI (XAI) using SHAP analysis for model transparency.
• Conducting disparate impact analysis to ensure gender-neutral model behavior.

Main Results:

• The platform integrates multiple ML models for comprehensive patient profiling.
• Explainable AI methods provide transparency into predictive model reasoning.
• Disparate impact analysis addresses ethical considerations and potential biases.
• The system aims to support tailored treatment planning for psychosis.

Conclusions:

• The developed ML platform shows promise as an experimental prototype for predicting MCT effectiveness.
• The integration of XAI and ethical bias analysis aligns with regulatory requirements (e.g., EU AI Act).
• This approach can advance personalized medicine in mental healthcare by supporting clinical decision-making.