How to change detection channel for cub-cellular detections?

Hi everyone,

Wondering if anyone has experience with the subcelluar detections in Qupath? Is there a way to change what channel the detection threshold is applied to? We’re hoping to find quantify hematoxalin stained spots if possible (see picture).


Many thanks,
Kris

This may already work if the channel you want to detect in was called anything other than hematoxylin: https://github.com/qupath/qupath/issues/304

At the time I wrote the command, I assumed that hematoxylin would be the stain used for nucleus detection and would not be needed for subsequent spot detection.

It sounds like this restriction within QuPath should perhaps be lifted… can you say a bit more about what is the use case for detecting hematoxylin stained spots?

Yep, make sure you are in Brightfield Other, as you cannot change the name of the Hematoxylin channel in H&E or HDAB. The other option is to copy and paste the H&E or HDAB color vectors so that the “Hematoyxlin” numbers are in Stain 2. Then you can perform the subcellular detection on the Eosin channel and find the hematoxylin spots.
image
Double click on the Image type to change.

@Research_Associate Thank you that’s perfect!

@petebankhead We are using Alkaline Phosphatase Blue conjugated probes to detect the presence of a viral DNA in the tissue. Figured the staining looked similar enough to hematoxalin to try that channel and so far it seems to have worked pretty well.

Thank you both for replying so quickly

1 Like

Ah, it probably isn’t exactly the same then, and you will almost certainly get better results if you perform the Analyze->Preprocessing->Estimate stain vectors, which can only be done in H&E or HDAB. Once you have those values, I would perform the swap or copy them over to Brightfield Other in some other order/naming scheme.

  • You can also just take the initial H&E stain vectors and swap them, moving the initially blue vector over to red, and the red over to blue, and manually align them that way, which lets you stay with the Image type as H&E and still perform the analysis. Might be easier in the long run.

Not exactly - it can only be done with two stains, but it should work with ‘Brightfield (Other)’ and different stain names.

1 Like

Great idea, I’ll give that a try thanks

Oops, you are right, I never tried Brightfield Other without editing the third stain, and it is only when I edit the third stain that I get locked out of Estimate stain vectors.