Microscopic Image Analysis for Cell Biology

Advances in microscopy technology are enabling cell biologists to design experiments which generate very large numbers of images of interesting biological phenomena. These images are sometimes two-dimensional but are often in three dimensions and are acquired over time, like 3D video. Such images need to be studied and analysed so that the results of the experiments can be interpreted. This is very labour intensive so there is a growing need for computer software that can analyse the images automatically. This is challenging because the images are highly variable even within a single type of experiment. Images also vary in scale from images taken of activity within a single cell to image sequences of entire moving populations of cells. In this project, computer scientists specialising in developing software to analyse images will work with life scientists to develop this automation. They will make use of recent developments in computer vision and machine learning which enable computers to learn from examples how to interpret images. Several lines of collaboration will be explored with an initial focus on analysing images of cell-based assays obtain using fluorescence microscopy. Software developed will be made available to researchers via a large open software project called OMERO.

Researchers from Computing will ?hop? discipline to the Life Sciences to initiate collaboration on automated image analysis for cell biology. Requirements for microscopy image analysis have shifted dramatically in recent years as technology developments have enabled higher throughput and higher resolution multi-dimensional, multi-spectral data sets to be routinely acquired, often at multiple scales. Algorithms that automate analysis of cell biology images are crucial as manual annotation of such data is often the rate-limiting step in microscopy workflow. The large data sets generated by high-throughput techniques enable modern machine learning methods to be leveraged: algorithms that scale well to large image data sets in which only relatively small numbers of images are partially annotated. Initial activity will focus on analysis of phenotypes from fluorescence microscopy images of in vivo cell-based assays. Software developed will be integrated and disseminated via the open OMERO project.