Cluster Analysis and Focus Stacking in Bioscience

One of the exciting things about research is that you get to try new things in new ways.  I am always amazed at how many “novel” techniques can end up benefitting science, but don’t because they are non-traditional or rarely used in a particular field.  The image above for instance is a traditional flat mount retina that has undergone a procedure to examine various aspects of retinal vasculature.  I will not go into greater detail here because it is work from a colleague that I’ve assisted with these data.  This collaborator was trying to figure out regions of retina and how to define them post treatment.  Their decisions were for the most part arbitrary, but I thought we could help them out a little bit by making more precise determinations of regions.  To do this we used standard cluster analysis approaches (that honestly have been around for a few decades) that helped tremendously with finding patterns of properties that emerged in these retinas.  I can’t say much more than that with these data.  We use cluster analysis in a slightly different way for our analysis with molecular probes as channels in our “multi/hyper-spectral” datasets, but one can also use different spectra or filters on tissues to obtain useful results.  For a nice tutorial written by Robert Marc on clustering aimed at anatomists click here.


The other nice tool that is surprisingly common now in some circles is focus stacking.  I’ve written about it before with for both histologic and non-histologic applications.  The results can be remarkably compelling, making optical images appear like colorized scanning electron microscopy images.  Since I’ve discussed those applications in some detail before, I’ll just note here that we’ve now tried an additional application with focus stacking in retinal vasculature for yet another collaborator and it works remarkably well.  I’ll wait until the manuscript is published to say any more than that, but the results are spectacular.


Leave a Reply

Your email address will not be published. Required fields are marked *