Masson's and Liver Lipids (The anti-stain?)


#1

Hi there,

Somehow I’ve come to be known as the tech savvy person in my lab, so they recently threw me into the deep end of the pool on using CellProfiler. For the first project I was successfully able to find and modify a pipeline to fit my needs. I was trying to quantify the area of glucagon in the pancreas, which was as simple as unmixing the stain colors. Then I did the same thing with insulin.

Now I’ve got a problem. I’m supposed to be finding the number of objects, and the area occupied by lipids in the liver. I thought I could handle it by unmixing colors again, but it’s Masson’s stain, which I can’t seem to get to cooperate. The biggest problem (to me) is that the lipids are identified mostly by the absence of stain. I tried masking the stained area, but that hasn’t seemed to work. I tried a custom filter, but couldn’t get that to work.

My boss sent me some powerpoint slides with previously analyzed images, thinking it would help me reverse-engineer a pipeline (the original creator of the pipeline is no longer in our lab, and I can’t get in touch with them to try and get it). The slides have only confused me more. I can’t figure out how they got the binary image, or how they got the “outlines” image to only select the lipid objects, and not the larger objects like blood vessels and the like.

Could someone just give me a nudge in the right direction? I can’t figure out what step I’m missing here.

Sample image:


#2

Okay, so I’ve fooled around some more, :dizzy_face: and I managed to make a binary image… but I can’t seem to translate the binary image into objects.

Once I can figure out how to identify the objects, I can do the math from there.

I’ve attached both of my pipeline attempts, both un-mixing, and binary.

Lipids WIP.cpproj (500.8 KB)
Lipids WIP-2.cpproj (404.0 KB)


#3

Hi there,

Can you have a look at these topics:
Masson’s trichrome liver fibrosis
Histology slide of adipose tissue

Although they are not directly related, but there’re some similar techniques you can use (unmix color with Masson trichrome, invert black/white, detect “empty” lipid droplets etc.) to extract your object-of-interest

I’ll return to you with a more specific solution a bit later.
For now, good luck and have fun !


#4

I looked at the articles, and the pipeline you made for the fibrosis, but I didn’t get very far. I guess at this point, I’m trying to avoid the unmixing option, partly because I wasn’t entirely sure which version of Masson’s they use in our Histology Services Lab, and every attempt I have made at figuring it out has just failed for me. But that may be something to do with what I’m trying to do with the unmixed images downstream that’s not working. But I don’t know.

I’ve got a new pipeline going, that seems to be working okay, except it’s still selecting objects that I don’t necessarily consider lipids. I’ve messed with smoothing, and object size range, but it’s still either selecting things I don’t want it to select, or not selecting things I want it to select.

Should I be unmixing somehow (masking something?), and then getting a threshold/inversion?
Lipids WIP-3.cpproj (424.4 KB)


#5

It might be helpful if you post an example image (preferably the original plus whatever unmixed/processed version you’re using in IdentifyPrimaryObjects), and annotate them to tell us which ones you think SHOULD be picked and which shouldn’t.


#6

Thanks for your help. My supervisor decided that Masson’s wasn’t the best option for this project after all, and I ended up getting images from ORO staining to work with instead. The last time they used Masson’s for lipids, they were working with obese mice, so they had large lipid droplets to work with.

The ORO stain was much simpler, but large portions of the images had bubbles, and I couldn’t quite get cell profiler to ignore them. I just worked around that by selecting the best area of example to work with that I could.

Thanks again!