Research
Deep-learning for outcomes prediction
We create neural network algorithms to learn clinical outcomes from genomics and histology. Our goal is to develop an integrated prediction frameworks that provide a holistic assessments of risk to improve clinical management.
Digital pathology software infrastructure
We develop open-source tools to support the management, analysis, and integration of digital pathology images. The Digital Slide Archive, HistomicsTK, and HistomicsML empower users to apply powerful image analysis and machine-learning algorithms to their digital pathology data.
Phenotype / genotype studies
Histology has been used to diagnose disease and predict outcomes for over a century. Datasets that combine digital pathology with genomics provide a window into the molecular foundations of histologic phenomena. We use quantitative image analysis to reveal and precisely measure these relationships.
