This research aims to study how multimodal data such as neuroimaging and EHR can be used to predict treatment and medication response for people suffering from Major Depressive Disorder (MDD).
It will explore the following research questions:
1. How and to what extent can machine learning (ML) approaches, utilizing EHR, demographics and imaging (fMRI) data, be used to to predict antidepressant treatment response in the MDD population?
2. How can multimodal and heterogeneous clinical data be combined and encoded to enhance machine learning-based prediction of treatment outcomes?
3. How do patient characteristics such as clinical history, demographics, or diagnoses shape model performance and support the identification of clinically actionable markers of treatment response?
Research objectives:
o1) To quantify antidepressant treatment response in individuals with MDD
o2) To investigate whether fMRI and other scales are associated with baseline symptom severity and treatment outcomes.
Scientific rational:
MDD is a major global health burden, yet antidepressant response varies greatly across individuals and is poorly understood in real-world settings. Traditional clinical trials lack generalizability, underscoring the need for large-scale observational data to characterize treatment outcomes and their predictors. UK Biobank integrates diverse mental health, clinical, and neuroimaging data, offering a unique opportunity to explore factors associated with treatment response and to inform personalized approaches to care for people with MDD.
Dissemination and compliance with UK Biobank’s AI policy:
Models will be developed and validated with interpretability in mind. We will publish our findings in peer-reviewed open-access journals, and make AI model code, trained weights, and relevant scripts available on GitHub or institutional repositories in accordance with UK Biobank’s AI policy.