Vascular stain identification


I am struggling to identify an endothelial marker (Claudin-5) in a mouse brain section of a stroke. There is a degree of background specifically in the stroke area when stained with Cl-5 using DAB. I am able to analyse the other areas of the brain easily enough, but I am finding difficulty with the area below as there is also staining of haemosiderin (yellow circular objects in the below image).
I have looked at a similar post relating to pig brain and hematoxylin, but the solution to that doesn’t quite work for us (Help in identifying vessels).
Would anybody be able to suggest an appropriate pipeline that would help identify the vascular stain only?

Thanks in advance


Since we are not familiar with the stain in question, would you be able to post a picture indicating specifically what features that want identified?

Also, aside from the above, the contrast between the regions of the image seems to be quite low: Distinguishing the yellow areas from the brown areas would be difficult, if at all possible. I tried googling Claudin 5 to see some examples, and saw one like this: … _id=242654. Here, the hue contrast is much more striking and would be much easier to detect. Is is possible to generates images similar to this one (at least in terms on contrast, if not the actual colors)?


Hi Mark,

Thanks for the reply. I have other areas of the brain which have stained very nicely and indicate the contrast of the Claudin 5 staining compared to the surrounding tissues very well as below, this image is from the same animal with all stained areas representing vessels with Claudin-5 markers. Essentially I would be looking to extract the same information from the original image as the second one.

I fear that the original image simply has too much background and has the complication of the haemosiderin deposits to analyse with any degree of accuracy. I also realise it will be difficult, if not impossible to differentiate between end-on viewed vessels in brown, and haemosiderin deposits in yellow.

Many thanks

Playing around with the image a bit, if it’s the dark brown regions you are looking for, I may have a suggestion…

Use ColorToGray to split the image up into the R,G and B components. It appears that the red channel preserves the featured you want, while reducing the influence of the yellow haemosiderin deposits. Use ImageMath to flip the pixel intensities, and then use IdentifyPrimaryObjects to find the now-bright regions.


Thanks Mark, that helped a lot! I will have a play around with the settings a bit for IdentifyPrimaryObjects, but the initial tests with it seem to identify the areas well.

Thank you very much for your assistance.

One other thing: If you are using Windows, you can use the pixel-based machine-learming tool called Ilastik that comes bundled with CellProfiler. Pixel-based methods of detecting regions of interest can sometimes work when more automated methods fail.

Basically, you would start up ilastik, give it some examples of images, and manually annotate the regions of interest vs. those that are not. ilastik will then produce an .h5 file which is used as input into the ClassifyPixels module (again, Windows only). ClassifyPixelswill then apply the classifier to each input image and produce a prodbaility map which can thresholded to yield foreground/background regions of what you want (hopefully). See here for some documentation on its use.