Imaging is used as a tool for discovery throughout the basic life sciences as well as in clinical research. Confocal, spinning disc, multiphoton, and light sheet microscopes are the new engines of discovery for many fields including cell biology, cancer research, and drug development. These advanced instruments produce large volumes of complex, multi-dimensional image data; data which may be crucial to characterize a cellular process or to identify a potential drug candidate. To make use of this microscopy data, we must transform raw pixel data into representations that scientists can quantify, manipulate, analyze, share with colleagues, and ultimately understand.
My research focuses on bioimage informatics, an emerging interdisciplinary field that deals with this problem. In the past, analysis of microscopic image data was largely performed by hand, a costly and error-prone endeavor. Our goal is to develop computational methods that automatically analyze and understand images of biological processes, using microscopic images as their primary source of information. We draw upon methods from computer vision, machine learning, statistics, and bioinformatics to quantify image data and to answer concrete biological questions. Accuracy and robustness are vital. The methods we develop must also be sensitive enough to recognize highly dynamic and complex events, and they must be scalable to handle very large volumes of data.