Lipid droplets in CellProfiler

Hello,
I’ve been trying to analyze lipid droplets in cells stained with Hoechst and Lipidspot (red channel), using the suggested pipeline on the website. However, when I observe the lipid droplets identification quality using edit objects manually, I see that not a lot of them are recognized when the guiding image is masked LD using nuclei objects.
I was wondering if anyone has any ideas about how to improve it?
Attaching all the relevant images and the pipeline here.


001__cntrl 1-3C_cntrl1a - try1.tif (12.0 MB)
pipeline try3.cpproj (919.4 KB)
Thanks a lot,
Rita

1 Like

Hello Rita,

By edit object, do you mean Thresholding Method?
In this attached pipeline, the RobustBackground have been used as the Thresholding method in LD detection, not the Manual method, correct?

Could you upload a couple of images, so I can run them through the pipeline?

Best,
Nasim

Hi Nasim,
Thanks for the reply.
I can now see that I did not attach the last version of the pipeline. In the latest version I added the module “EditObjectsManually” after the “IdentifyPrimaryObjects” in which I identify the lipid droplets, so that I can see more easily the quality of the identification. Regarding the thresholding method , you are correct. In the second “IdentifyPrimaryObjects”, I used Robustbackground, as was suggested in the tutorial on how to identify speckles, however, it didn’t seem to work.
How many images do you need to run the pipeline? Will two be enough?
Thanks again
Rita

pipeline try 4.cpproj (115.3 KB)
Cntrl1c.tif (12.0 MB)
EP1c.tif (12.0 MB)

Hi Nasim,
The images that are shown in the pipeline are from the same batch as the images I uploaded, so I tried to analyze them all in the same way.
Regarding the metadata, sometimes I have issues with it too. What I usually do is switch in Images to “no filter”.
Regarding your question, what I’m trying to measure are the speckles in the red channel, but the issue is that they are not being identified very well, even though I followed the tutorial instructions found on CellProfiller. It seems that trying to adjust the typical diameter is not enough in my case.Moreover, It seems like the areas being identified does not contain speckles.
Thanks for the help,
Rita

@R_SH,

Where did you find the suggested pipeline for lipid droplets on the Cell Profiler website? I have been trying to make my own pipeline for a while now with no luck.

Best,
Ben

Hi Ben,
It’s the suggested pipeline for speckles, not specifically for LD. Unfortunately it does not seem to work for me either.

Best,
Rita

Hi Rita,

32-bit images, such as yours could throw off some of the thresholding. Would you need 32-bit images?

Please take a look at the attached pipeline. I wrote some comments in the summary box on the top right. I wasn’t sure if you wanted masking to detect the lipid droplets only in the nuclei, so there are two “IdentifyPrimaryObjects” and two “EditObjectsManually” modules for you to compare (unmasked vs masked).

Back to your original question, what you see in the “EditObjectsManually” are the detected objects from the previous module (“IdentifyPrimaryObjects”) and using a masked image as a guide will only limit the observed area to the masked image area.
For detection of objects, choosing the appropriate thresholding method and finding the best thresholding limits are very important. You can read more about different methods from the help menu shown by “?” button.

Let me know if you have any questions.

Sincerely,
Nasim
pipelinetry4_2.cpproj (1.5 MB)

Hi Nasim,

Thank you so much for your help! Unfortunately, I can not open the pipeline you attached, it shows me that there has been an error.
Is it possible to try and upload it again?
Best wishes,
Rita

pipelinetry4_2.cpproj (1.5 MB)

Hi Nasim,
Still does not work, could it be that you are using mac and I am using windows?
Anyway,attaching the error.
Maybe a word file with a print screen could solve the issue?


Many thanks,
Rita

Hi Rita,

I downloaded the pipeline again and it opens ok here. There seem to be an unreadable folder name (here shows with ??? ???) on your desktop. Could you try putting the pipeline on your desktop, not in that folder?

I also attached a pipeline version (not a project) here. After opening the pipeline, upload the images you sent me (by drag and dropping them to the specified area) and give it a try.

Best,
Nasim
LipidDroplets.cppipe (28.2 KB)

1 Like

Hi Nasim,
It works!!! Thank you so much! you are a real lifesaver!!
If I may ask about the intensity re-scaling, is it to re scale it to 16-bit? and how do you decide the pixel numbers?

Again, thanks a lot
Rita

Glad to hear that Rita and you’re most welcome!
Yes, the RescaleIntensity module is to scale it down (in this pipeline to 16-bit).

By pixel numbers, do you mean bit-depth? If yes, it refers to the binary range of possible values (shades of grey or colors) in an image. For example, the most popular 8-bit-depth image have a binary range of 256 (2 x 10^8) possible values for a pixel, and a 16-bit image has 65,536 (2 x 10^16) possible values.
Depends on the sample contrast and desired image resolution, one can choose the bit-depth. If the sample has high contrast, 8-bit image would be sufficient.

Here are a couple of links to read and learn more:
blog post on top 10 tips for image acquisition
Basics of Bioimage Analysis

Best,
Nasim