Clustering cell morphology after segmentation of image stacks


Hi all: I’m sure this has been done but we’ve been banging our heads without success. We have used segmentation and 3D Manager to measure the shapes of genetically labeled (conditional fluorescent protein) cells in a tissue (confocal image stacks). This part has gone fine, we have various parameters from 3D Manager (volume, area, feret, etc) for each of the segmented cells. Now we would like to find ways to cluster the results to determine whether the individual cell shapes fall into clusters (for instance, large vs. small, branched vs. spherical, large and branched vs.small and spherical, etc). We have tried for months to use affinity propagation clustering (in R), but the clusters it gives us do not make sense, regardless of which parameters we use/exclude or how we weight or normalize the distribution. So, like I said, surely someone has done this already, I was wondering whether someone could point us in the right direction, either in the form of a plugin or series of plugins, or a manuscript on this topic?

Thanks a lot for your help. Howard



Have you looked into Bio7 at all? There are so many goodies in that package, and it’s compatible with R. Perhaps you will find something there?



Let’s have a look! Thanks etarena.