Is this data usable?

My master’s project revolves around getting a 3d image of the trigeminal ganglion that’s stained with NeuN. Currently I’m struggling to find good software to analyse my images. All the examples I see online are using perfect Z-stacks so when I try to replicate it on my data it doesn’t work. I’m just wondering if this can even be counted with automated methods?


This is a maximum projection of one of my samples.
I tried ImageJ 3D object counter but the thresholding is too simplistic and excludes a lot of my cells while the algorithm will extend and mark whole areas as a cell because there is one pixel connecting it.
The best thing that kind of worked so far is QuPath but it isn’t really equipped to deal with Z stacks well. I’ve also tried CellProfiler.
No one in the lab has done such analysis before so I’m looking for help here…

Do you mind elaborating a bit more on what kind of analysis/measurement you want to do? It looks like some regions may not have enough information to segment individual cells but maybe it is easier to see in 3D than this projection, and maybe you don’t even want to segment individual cells?

And can you post an actual 3D stack here or online somewhere?

1 Like

I vaguely remembered this…


Maybe it would be worth giving a try?

Unfortunately, I have been using Imaris for 3D segmentation, which is not very open sourcey, and very expensivey. So don’t really have much else in the way of suggestions :slight_smile: Good luck.

2 Likes

Thanks for replying. Yes, so I’m trying to segment/count individual cells as I’ll also be staining the same structure for GFP (we have a transgenic mouse strain) and then try to see how many neurons express GFP compared to the total number of neurons (NeuN).

I’ve included 2 tif files - 2 Z stacks.
But as you were saying I think some regions are hard to segment…

1 Like

Thanks. We actually do have Imaris at our bioimaging computers so I might go down there to test it out! Did you find it the best out of the software you used so far for 3D segmentation?

1 Like

Hi @Mancunian,

@AnneCarpenter

I checked your images. Yes as mentioned in the previous discussions, in many regions of the image the information is very less. I am not sure about your cell boundary, but since your interest is to get the count of the cells, you can get the count by segmenting the Nuclei. You can try Nuclei segmentation after applying may be Median filter. Just to start with you could explore this example 3D segmentation in CellProfiler.
Hope this helps you to start with.

Regards,
Regards,
Fujifilm Wako Automation (Consultant)
www.wakoautomation.com
For CellProfiler training or optimised pipeline write to,
lakshmi.balasubramanian.contractor@fujifilm.com

Join us at the High Content Analysis Symposium Oct 21-22, London
Society for Laboratory Automation and Screening

SLAS 2019 Advanced 3D Human Models and High-Content Analysis Symposium

Read more on our site.

Yokogawa CV8000 - The Ultimate in Confocal HCS
https://www.wakoautomation.com/products/yokogawa-high-content-imaging

1 Like

Hi, yes I tried following the tutorial you posted but I think I encounter problems because not all the cells are give the same amount of fluorescence, meaning I lose cells if I lower the threshold and get too much noise if I increase it.

It has been fairly good, though it is not without it’s flaws. It is also the only 3D segmentation software I have used, I am fairly new to third and fourth dimensional data sets.

1 Like

@Mancunian,
Yes. I got it. But we also should understand the fact that as you had already mentioned about the uneven fluorescence in your images. So the during thresholding there would be problem because your intensities might not fall in the given range. So you might have to get rid this problem first as I had mentioned earlier as much as we can so that segmentation becomes better. This is true with any image analysis software.

I would like to add that, any software that is giving the result (segmentation) is mostly based on our image, so segmentation result is based on our image Signal/Noise ratio. Even by filtering/Enhancing/Suppress steps we are trying to make it better, but sometimes we may not able to get all the complete details which could be our limitation.

The example I shared is to start with. But I am not sure if you tried applying some filter (for example Median filter) to get your intensity range smoother in your image. Do you mind sharing your pipeline so that we could help you better?

Regards,
Fujifilm Wako Automation (Consultant)
www.wakoautomation.com
For CellProfiler training or optimised pipeline write to,
lakshmi.balasubramanian.contractor@fujifilm.com

Join us at the High Content Analysis Symposium Oct 21-22, London
Society for Laboratory Automation and Screening

SLAS 2019 Advanced 3D Human Models and High-Content Analysis Symposium

Read more on our site.

Yokogawa CV8000 - The Ultimate in Confocal HCS
https://www.wakoautomation.com/products/yokogawa-high-content-imaging