color adjustment of histology slides

[Rasaa84]

Hi,

I'm looking to modify the colors of histology slides to match a reference image or a specific section. I am wondering how I can do this using pyvips?

[jcupitt]

Hi again, there are loads of ways of doing this, you'd need to be more specific about what you want. Do you have some sample images and an expected result?

[Rasaa84]

This is the input image:

I need to change its color to the color of this image:

[jcupitt]

That looks like a SVS image from openslide compared to an SVS image from Leica's viewer, is that right?

You need openslide4 plus libvips 8.15, it has all this built in (phew).

[jcupitt]

For example:

That's an SVS slide with and without colour management. Colour management is a new feature of openslide4 and is supported by libvips:

https://github.com/openslide/openslide/releases/tag/v4.0.0

[Rasaa84]

Yes, it's right. one is a SVS from Leica's viewer and the other from openslide. Thanks John for your help. I am wondering if I can use color management along with dzsave?

[jcupitt]

• I'd remove orc and add hwy, assuming you have a recent enough version available.
• You could also swap libpng for libspng.
• You have librsvg and orc in there twice.
• You don't need libgif
• I'd add libcgif, if available

You have untested libraries in there, so you are very vulnerable to attacks. I would not use this libvips to process any user data.

[Rasaa84]

Regarding untested libraries, which libraries do you mean specifically? probably I can remove them.

I have also another question. I have used the same API to fix the color for two different image from different scanners. The outputs are as follows:

As you see the colors are different. The API changes the color of image from Leica (left image) while keeping the color of image from other scanner (right image) unchanged. I'm interested in obtaining uniform output colors. I'm specifically interested in harmonizing the slide stains to match the color of the left image, aiming for a more pinkish tone rather than a purplish one. I know that there are different approaches for stain transformation but I am wondering if libvips provides such possibility.

[jcupitt]

Regarding untested libraries, which libraries do you mean specifically? probably I can remove them.

libmatio is probably the worst -- it hasn't been designed for untrusted input, so it is very easy to attack. Unless you really need to load Matlab save files, I would remove it.

If you are loading internet data, I would remove all libraries that you don't need.

I have used the same API to fix the color for two different image from different scanners.

You'd need to look at your set of scanners, see which ones include an ICC profile, and check if openslide supports ICC profiles from that scanner.

But that probably won't be enough to get a good match. Many scanner manufacturers use ICC profiles to make their slides images look "better" (ie. brighter, more contrasty, more colourful), and have no interest in colour accuracy.

The best solution is to use a scanner target and to calibrate the devices yourself, eg.:

https://www.appliedimage.com/applied-image-announces-nist-traceable-color-transmission-calibration-slide-digital-pathology-imaging/

(just the fist hit on google, there are many of these)

[Rasaa84]

John, Thank you so much! Appreciate your help.

[jcupitt]

It's pretty easy to use libvips to find a colour correction matrix from a scan of a calibration target, if you end up going that way.