Histological slide analysis of atherosclerotic plaques

Hello everyone,

I’m a grad student that is currently trying to (partly) automate the image analysis of our histological slides of atherosclerotic plaques as a side project. I’m very new to this and I’m a bit overwhelmed by all the different options that are available when it comes to software. I had a few questions regarding the image analysis of plaques in the aortic valve:

  • At the moment, I’m exploring with simple pipelines in Cellprofiler, would this also be the software you would recommend when looking at histological slides at this magnification?

  • I was currently thinking of outlining the valve itself and the plaque tissue of the valve as a whole and the three subvalves manually, as I couldn’t see a way this could be done automatically. Do you think there might be training software to do so automatically? And more generally, which software/training algorithm would you recommend using for histological slides when you’re not necessarily interested in every individual cell?

  • The white crystal-like structures, alot of them can be seen in the lower right corner of the plaque, are as white as the background but they are part of the plaque area. Is there a way to distinguish background from these cholesterol crystals?

The information I want to obtain in the end is the percentage of the area of each plaque that is stained, if possible per subvalve. As it may be important for training, I’d like to add that although the picture below was still taken manually with a microscope, we do have a slidescanner now that is able to make much more consistent and high res scans (.tiff).

Any help would be greatly appreciated!

1 Like

Hi @PepR,

In CellProfiler you can try using “Unmix” colors module. Basd on your stain you can split the colors & use the resulting image for further processsing. There are tutorials & example pipelinse available.

The other option could exlore the Qupath software which would be helpful for histopath images.

Regards,
Lakshmi
www.wakoautomation.com

thank you for your quick answer @lakshmi

I was trying to use EditObjectsManually to define my the border of the aortic valve, but it seems to be very slow on cell profiler, so I was thinking about maybe doing that step in ImageJ first. I wondered if you knew whether cellprofiler can recognize a region by an outline that is already present in the image (e.g. red outline = whole valve, green outline = subvalve 1, etc.),so that it only performs the analysis within these defined outlines.

In case you try QuPath, with the Thresholder I was able to get this in about a minute.


Obviously I don’t know what I am drawing or what the correct threshold would be.

The crystals look like they would be more difficult. You could try a brightness AND size/major vs minor axis classifier to isolate them from the lipid holes and other tissue tearing already present, but I wouldn’t count on amazing success without some sort of context based classifier.

2 Likes

Slight addition, CellProfiler might have better options to identify the crystals, I am not sure, but one option would be combining the two softwares.

QuPath can easily handle the whole slide images you mentioned, and has great annotation tools. It also has the ability to export the smaller regions corresponding to those annotations for later analysis in CellProfiler.

2 Likes

Hi @PepR,

O.k. I am not sure how you tried the EditObjectsManually because generally it doesn’t take time. In case you are able to share the pipeline we could help you better.
Anyways, I tried the pipeline without manual intervention. It looks ok but need some optimization. Also I am not sure is these are the regions you are looking to segment.
Here are the screenshot,
For your green valve circled in the image

for the larger red region (Based on the region you are looking for you can play around with the threshold settings

for Crystals (can be optimized)

Please find the pipeline
histopath_seg_plaque.cpproj (403.7 KB)

As @Research_Associate pointed out you could use the combination of both these tools too…

Hope this helps,

Regards,
Lakshmi
www.wakoautomation.com

1 Like