Transforming microscopic and medical imaging into powerful quantitative tools with bioimage informatics

Since its invention, the microscope has played an increasingly important role in the life sciences and in clinical research. Today, devices such as confocal, multiphoton, super-resolution, and light sheet microscopes have become the new engines of discovery for many fields including cell biology, neuroscience, and drug development. A similar trend can be observed in the clinic, as physicians are increasingly reliant on x-ray images, MRI, and tissue scanners for screening and diagnosis.

While historically, the analysis of microscopic images has been largely qualitative, technological advances and the coming of the digital age have pushed a transition towards quantitative analysis. Modern microscopes and medical imaging systems routinely produce enormous volumes of complex, multi-dimensional image data – data which may contain information crucial to characterize a cellular process, to identify a promising new drug, or to diagnose a cancer patient. To make use of this data, we require reliable automatic methods to transform raw pixels into representations that we can quantify, analyze, and understand.

My research focuses on the development of computational models to quantify and analyze biomedical images. In this talk, I will give an overview on methods adapted from computer vision and machine learning which promise to dramatically improve our ability to understand and interpret biomedical image data. I will touch on topics from my own research including methods to efficiently search and uncover rare cell behaviors in large datasets, and to automatically recognize breast cancer using deep neural networks.