Interpretability and model auditing

I develop methods to understand what predictive models learn and how their information compares with expert knowledge, including Signature Activation and Shared Canonical Component Removal.

Cardiovascular imaging

I use machine learning to extract measurable features from coronary angiography and study their relationship with disease, including coronary artery tortuosity.

Precision psychiatry

I study factors associated with antidepressant treatment dropout and investigate what predictive models can reveal beyond expert-defined clinical features.