We develop machine learning and computer vision methods to learn meaningful representations, generate images, and model how complex systems change over time. Our research focuses on generative models, representation learning, physics-informed learning, and spatiotemporal modeling.
Healthcare and biomedical discovery are central to our work. Challenges in medical imaging and physiology motivate new learning methods, which we study through algorithm development and biomedical evaluation. Our goal is to support diagnosis, disease monitoring, and treatment planning with reliable and interpretable AI.
Our work connects methodological questions in ML/CV with clinical needs, with publications in venues including NeurIPS, ICLR, CVPR, ICCV, ECCV, MICCAI, IPMI, and IEEE TMI.
Research Directions
We connect advances in learning and vision with questions in biomedical imaging and healthcare. Student projects are developed collaboratively around methodological interests, technical background, and biomedical relevance.
Machine Learning & Computer Vision
- Generative Models
- Representation Learning
- Physics-informed Learning
- Spatiotemporal Modeling
Biomedical Image Computing
- Multimodal Learning
- Image/Video Generation
- Longitudinal Prediction
- Physiological Modeling
Clinical Applications
- Brain Diseases
- Structural and Functional MRI
- Cerebrovascular Flow Modeling
- Disease Progression & Treatment