research

Our research focuses on addressing the challenges of translating machine learning (ML) approaches into clinical practice through three interconnected themes. Each theme is grounded in real-world clinical applications.

Integration of domain knowledge. How can we effectively incorporate the rich, multi-modal biomedical knowledge found in structured sources (e.g., knowledge graphs) and unstructured text into ML algorithms? Our goal is to leverage this domain knowledge to develop robust, generalizable decision-support systems that align with clinical needs.

Learning with limited annotated data. Given the scarcity of annotated clinical data, what strategies can enable the development of high-performing ML models? We explore approaches such as self-supervised learning and the integration of domain expertise to create clinically useful foundation models that function effectively even with limited data.

Responsible machine learning. How can we design and evaluate ML systems that promote equitable healthcare for all populations? Our work focuses on developing and implementing human-in-the-loop systems to augment medical decision-making while emphasizing fairness, transparency, and patient safety.