Evaluating methods to identify cells in fluorescence microscopy images

Developing efficient and accurate methods is an important aspect of conducting research. A common technique in cell biology is to visualize and quantify proteins in cells using fluorescence microscopy (Snap 2001). The accurate and efficient quantification of proteins, particularly for large numbers of microscopy images, requires a computer based analysis. There are two broad types of methods: using fluorescent cells and using non-fluorescent cells. I propose to evaluate the performance of these methods in two ways: determining how accurate the segmentation is, and evaluating what impact the accuracy has on estimates of gene expression under conditions where cell morphology is changing. The results will help inform cost-benefit analysis of different segmentation approaches.