|
Abstract:
|
In biomedical research, it is often necessary to study cell population characteristics, and quantify changes in cell phenotypes on a cell-by-cell basis. Traditionally, this work has been performed by interactive manual use of a microscope. In disciplines like systems biology, studying topologies and dynamics of complex functional networks of cells, massive systematical screens for phenotypic changes in cell populations are required. Also in drug discovery, effects of pharmacological agents on the populations must be tested automatically in a high throughput fashion.
The development of robotic arrayers and automated microscopes, together with increasing computing power and storage space have enabled the automated screening of cell populations, resulting in a revolution of microscopy imaging. Currently, imaging of hundreds of populations in parallel is common practice in a single experiment. During the screen, images of each of the cell populations are stored for subsequent analysis. The amount of image data renders manual visual analysis impossible, requiring automated image analysis systems, and software.
Current procedures of automated analysis in high throughput microscopy, however, have several drawbacks. Standard practices exist for a number of analysis approaches, but especially three dimensional studies are generally performed manually, or semi-automatically. Furthermore, greater care must be taken on practical issues, such as low computational cost and easy implementation to advance routine high throughput screening studies by bioscientist. This thesis considers fully automated methods ranging from cell enumeration, to subcellular analysis in two and three dimensions, concentrating on the applicability of the algorithms for high throughput microscopy. |