Segmentation of Ramified Cells

segmentation
cells

#1

Hi,

I am a new user eager to learn as much as I can. I need special help with the processing of ramified cells. In all the tutorials I have seen the examples are posted with convenient round cells. However, I am working with astrocytes, neurons and microglia, which a highly ramified cells. I would like to analyze morphological data from this kind of cells. The first problem I have is the segmentation of the cells as shown in the picture below. I imagine a computer cannot recognize the numbers of cells as the eye can do, but I am wondering if there is some way of clustering the different objects identified by some parameters such as distance. I really appreciate the support you can give me in this issue


#2

Hi,
Can you describe in a little more detail what you want and what’s not working well? Show us an area of an incorrect segmentation and describe what you think the correct segmentation should be? It’s hard to help without knowing more detail.

That being said, segmentation of round objects IS certainly easier than ramified ones- is it feasible to include in your assay a nuclear or soma-only stain, even just as a first step? That may be more successful, but again, a lot depends on what you’re trying to extract from these images.


#3

Dear @bcimini,

I appreciate your response. So, I need to take out the morphological data of cells imaged in 3 channels. One channel is Iba1 (microglia cells), CD45 (immune cells) and DAPI (nuclei of all cells). So, I have pictures from two different brain regions, striatum and cortex (I pre-processed in imagej) and I need to characterize and further classify the Iba1 and CD45 cells in cellprofiler analizer. In the striatum, cells are denser and the shape is more circular, while in cortex cells are highly ramified. So, I build two different pipelines. For ramified cells, I must run “splitormergeobjects” Module to reconstruct the fine pieces, while for round cells I avoid this module (I attach pipelines). I did not use any nuclear staining in the process because I do not know how this can help. But I would appreciate your feedback. My other problem is that for some reason I am not able to make a database for processing in cells analyzer. I would really appreciate if you can 1) help me to realize if it is really necessary to have two pipelines and/or to reach better segmentation. 2) improve the pipeline and generate am adecuate database for the analyzer. I also attach a link for a couple of sample pictures.

Sample pictures: https://drive.google.com/file/d/1VzTywXeXOh5HllLV-C7haxO-ye-tVnj7/view?usp=sharing

Morphological Analysis-Cortex.cppipe (14.1 KB)
Morphological Analysis-Striatum.cpproj (519.4 KB)