Cell detection for Inform images

Continuing the discussion from How to create channels in a composite.tiff image in QuPath, when the image was obtained in inForm:

Hi,
I had the exact problem as yours with Inform images. Thanks to your tip about component_data.tiff,
I am now able to open the tiff file but cell detection is not working.
Were you able to resolve your issue?

I assume this is still about QuPath? Adding this post to the QuPath subforum since it was not placed there.

You also have not included which of the suggested steps you tried, and what error messages, if any, you were getting in the Log (View-Show log).

Thanks for your prompt reply!
I am not able to do cell detection in QuPath with Inform (by Akoya/Perkin Elmer) generated TIFF fluorescence image.
I am a newbie to QuPath and I havent used any imaging software other than Inform.
The steps I did were as follows:
Added image to QuPath. I am able to see all the channels when I click on the Brightness/contrast button so I guess this step is correct.
I created a Project and I am able to see the project name above the image list.
Next, I chose a rectangle tool to select a region in the image.
Then Analyze>Cell Detection>Cell detection.
Detection Channel is DAPI. I didnt change any default parameters that came up.
When I hit Run, I didnt see any changes.

The message I get in View log is:
INFO:0 nuclei detected
Processing complet in 0.08 seconds
Completed!

Thanks!

Same answer as previous post.

Whole image annotation is under Objects->Annotations. You need some way to tell QuPath what to generate the cells inside of, even if that is the whole image.

Whoops, my fault. You did say that you created a rectangle. Next, have you checked the Cell detection tutorial to make sure your Threshold is right, and that the pixel size makes sense?

That is probably going to be the issue. I can’t tell you which parameters are the problem just from knowing that no cells are detected though.

This may be useful as well - it is a bit old but applies more specifically to fluorescence.

The multiplex section of the documents also probably applies more specifically to your situation.
https://qupath.readthedocs.io/en/latest/docs/tutorials/multiplex_analysis.html#detect-measure-cells

Thanks for the links, I will check them out.

Can you elaborate on what you meant by “default pixel size “makes sense,” as shown in the Image tab”?
My image mag is 20x and pixel width is 1392 px and ht is 1040 px.

Pixel width should be something less than 1 at 20x. Those numbers sound like your image width and height, which is not relevant here.

If the pixel sizes are between 0 and 1, you are probably getting the metadata read correctly. You mostly want to look at the threshold in that case, as mentioned in the first link.

If they are unknown - that will make cell detection much harder as almost none of the default values will work.

Oh! sorry pixel width/ht is less than 1um.

You’ll usually have to make changes, especially for non-brightfield images.

The one I think of first is the Threshold option. The default is 100, but a suitable value here may be closer to 1 or 2 – I suggest trying a few.

You can get a feeling for the range of values in the image by checking the numbers in the bottom right of the QuPath viewer. For brightfield images these are normally in a similar range (so defaults work for detection), but for other images they can be very different indeed.

Reducing the threshold worked. The youtube videos no.4 and 5 were very useful to understand what each of the parameters do-thanks!

2 Likes

Hej @SD_ar I had given up a bit, and continued analysing only using INFORM. But I just retried following the links above and it got better! Which you tubes video are you referring? It would be nice not to depend on INFORM so much!

Hi Laia,
Yes, there is only so much you can do with Inform. That said, it much much easier to use.
I am not finding it easy to use Qupath. And I am finding that nucleus detection for my image is far accurate in Inform than with Qupath. It is very likely that I am not doing it right. So I am leaning toward using Inform.
Let me know how easy it was for you to use QuPath and what helped.
Regarding the videos-in QuPath I went to Help>Youtube channel (web). You will find several Pete’s videos.
I learn better with videos than written material:)