"I believe Digital Pathology will become the next big ‘OMICS’ that will revolutionize the way we extract information from patient tissues, that should help us make better and faster treatment decisions."
My driving interest has been in understanding and quantifying biomedical images through various computational feature extraction techniques. I joined Genentech in 2008 as a postdoc in the Weimer lab, in the department of the Biomedical Imaging. There, I explored the relationship between microglia morphology, gene expression and behavior in the living mouse brain, by analyzing data from 2-photon imaging and histology. Afterwards I spent 7 years in the pre-clinical Safety Assessment department, developing image analysis algorithms to quantitatively characterize the extent of toxicities observed in animals after experimental drug treatment. In 2018 I joined the Oncology Biomarker Group, and started building up a digital pathology group with the focus of applying advanced computer vision / deep learning methods to extract information from patient tumors based on histology, to enable better treatment decisions for our patients.
J Pathol. 2019 Nov;249(3):286-294. doi: 10.1002/path.5331. Epub 2019 Sep 3.
Nat Commun. 2020 Nov 4;11(1):5583. doi: 10.1038/s41467-020-19408-2.
The focus of our group is building digital pathology algorithms to efficiently extract tissue information from patient tumor tissues. We utilize deep-learning based methods as well as other computer vision and image analysis techniques to identify the primary features of images that correlate to clinical endpoints. A particular interest area is using multiple different strategies to not only predict patient outcomes from images, but making those prediction explainable using deep learning network visualization, and object segmentation techniques. By elucidating the biological mechanism behind our predictions, I believe we can build more robust tools to inform drug development strategies.