Cell Segmentation for Blotchy DAPI

I am trying to get cell segmentation on a 20 um tissue slice mounted on glass slide. The tissue is placenta which has a lot of connective tissue and non-uniform nuclei (variations in size and shape). Images taken with a spinning disk confocal.

I’ve been unsuccessful using a standard cell segmentation protocol in Fiji (Find maxima: segmented particles). Thresholding or max. filtering makes the nuclei large and bleed into each other which makes it even more difficult for the software to segment properly. Does anyone have experience with dealing with these kinds of samples?

Attached is a snip of what my tissue looks like in Fiji and the TIF file. Thank you!

example of DAPI.tif (8.2 MB)


What is the end goal of this analysis? If you just want to segment the DAPI staining… should be straightforward with the signal you have. If you want to ‘count’ those nuclei or segment one from the other - that is obviously more tricky. If your eye is having trouble distinguishing between nuclei - a computer won’t necessarily have an easier time. Trying to split nuclei here too that aren’t so circular using Watershed - for example - will also cause problems. This is an issue with tissue… (ha! sounds funny actually!)

So - let’s start with what your end-goal in this analysis is… that will help dictate which tool to use.

If you solely want to count cells… you can try out TrackMate’s spot detector. I’ve used it for bacteria (easier to id though) in tissue and it worked great. You just need to know a size estimate … obviously, you aren’t tracking these cells - but @tinevez’s spot detector is awesome. So use it just for that - to find your nuclei.

If you have the option or necessity to acquire more samples in future, it would also be a good idea to keep an eye on saturation, tip 2 here for reference, although the whole list is gold.

1 Like

Hi @etadobson , thanks for the response.

My end goal is to compare gene expression between tissue samples. Because the samples are from different tissue preps, the tissue section sizes will be different. I thought I would do cell segmentation or count nuclei or something to be able to correct for size difference before I compare the gene expression (another problem for another day). This was my thought, although I’m willing to try another method if you have one!

I will play around with TrackMate and see what it can do for me. Thanks!

This is a great article @awant, thanks for sharing!