In collaboration with the Imaging Platform of the Broad Institute, we develop deep-learning-based methods for detecting and measuring the phenotypes of cells in biomedical images. This includes cell segmentation, feature embeddings, phenotype classification, and image synthesis, with applications in drug discovery and functional genomics.
Generative modeling of imaging data
Modern deep learning architectures used for image analysis are often black boxes that cannot be interpreted in a direct and clear way. We aim to harness the power of generative models to provide meaningful interpretations of visual differences between biological images and to correct for nuisance factors of variation that could confound the correct understanding of biological signals.
Decoding spatial cell organization
After single cell biology, the next frontier in basic biological research is to crack open the mysteries of how cells work together to form higher level living structures, such as tissues and organs. Highly multiplexed imaging techniques that capture tissue images with single cell resolution are useful for studying the microenvironment of cells and the way they interact and communicate. We aim to use state-of-the-art machine learning and computer vision techniques to systematically analyze these images in order to uncover the patterns that govern the collective behavior of cells.