Research
Our research develops machine learning and image analysis methods with emphasis on limited supervision, interpretability, and real-world impact across healthcare, public health, and scientific imaging.
Limited Supervision
Methods for training deep networks when annotated data is scarce, expensive, or unreliable, including semi-supervised, self-supervised, and weakly supervised learning.
Semi-supervised learningSelf-supervised learningContrastive learning
Healthcare Imaging
Machine learning for radiology, pathology, and cardiology with emphasis on clinically meaningful prediction, localization, and explainability.
Semi-supervised learningSelf-supervised learningWeak supervision
Scientific Imaging
Machine learning and image analysis for microscopy from nuclear-materials to connectomics. Converting dense ultrastructural EM into segmentations and reconstructions for neural circuit analysis.
Image segmentationSelf-supervised learningMorphology analysis