Dear community,
In my bachelor thesis I am looking into 4 channel IF stainings of the mouse hippocampus. Until now I had to manually count the marker positive cells in the stacks using Fijis cell counter, often amounting to several hundred cells and hours of work.
To shorten this tedious process I want to know if there is a (preferably open source) software to deal with automatic cell counting in a z Stack.
As the morphology of the markers are very different (Radial glia processes vs nuclear stainings) I dont expect that everything can be automated. But it would help me greatly if BrdU, MCM2 and beta-galactosidase could be semi automatically assessed/segmented.
Several challenges that arise from the dataset are the very close proximity of the b-gal+ cells as well as their appearance on different slices because of their size. Furthermore the staining intensity is less towards the middle of the stack due to antibody penetration.
Are there any pipelines or programs to deal with this kind of dataset? CellProfiler only seems to be working on 2D images and neither Icy nor Fiji showed my any solution "out of the box". I would be willing to adapt current plugins to make them work on my dataset, but I am not sure where to start.
Perhaps some expert in image data analysis is reading this and can help in this matter!
Attached I provide a view of the different channels in a typical image as a Maximum intensity projection. If you require raw data for testing an idea, I will happily provide it.
Thank you all for your suggestions and your help!
Best regards
Paul
Dear Paul Wagner , you can:
- assemble a pipeline in CellProfiler with 3D modules; it will help you for running in batch mode large amount of images.
- use 3D ImageJ suite which relies on 3D ROI manager (macro recordable); NEUBIAS Academy webinar held by Thomas Boudier https://www.youtube.com/watch?v=OPC2kP-5By4 .
- use ZELDA plugin for napari [conflict of interest: I am the author]; it uses a simple protocol for 3D segmentation and visualization in napari, see more at: https://github.com/RoccoDAnt/napari-zelda .
Hope it helps,
best regards,
Rocco
Dear Rocco D'Antuono ,
thank your for your reply with showing different options! Because I lack experience in CellProfiler, either ImageJ or Napari will be my way to go.
Installing the ZELDA plugin was very easy and I am familiar with the basics of Watershedding/Threshholding.
Do you have any recommendations for starting settings or could take a look at my dataset? Here is a link to a downloadable version: https://drive.google.com/file/d/13BC24NJ_VgKyAbR-UN3j4DGtJc2ccR2W/view?usp=sharing
The Otsu algorithm produces a decent segmentation for everything with "medium intensities" but seems to struggle if the signal is concentrated in strong Puncta. Have you also implemented the Yen algorithm in the plugin?
Thank you for the help and building this useful toolkit for everyone to use!
Best regards,
Paul
Dear Paul Wagner , thanks for checking the ZELDA plugin for napari! At the moment, only the Otsu threshold or manual one are implemented. For puncta, I guess you would need to use the manual slider (below the Otsu checkbox) to see which value works for the segmentation.
I had a look at the data set: please remember to split it into individual channels; once obtained the labelled objects you can measure them in the channel you want. For example, you might use a nuclear staining to find the cells and measure those objects in the BrdU channel to see which ones are proliferating.
A paper describing the plugin will soon e available:
https://www.frontiersin.org/articles/10.3389/fcomp.2021.796117/abstract
Best wishes with your analysis,
Rocco
Dear Paul,
I would also recommend Cellprofiler, it is easy and there are tons of tutorials around. If you have problems with cell segmentation I would try Ilastik, which is also open-source and pretty handy. You can train Ilastik just by marking the right cells in a sample stack of images and it will recognize them in whatever other image you hand over to it.
Best, Caro
- Badges
- Science method