What is the significance of the parameters qupath detects about cells

What is the meaning of the parameters detected by qupath’s cells, and what is the design based on?For example, cell:Max caliper represents the value of the cell?What does this mean for cell classification?Thanks a lot!

As a general suggestion, most of the values can be searched for. Googling “min and max caliper of objects” gave me:

Hi @day0,

I’m not sure exactly what you want to know, but if I assume this screenshot is from the detection measurement list, generated after a cell detection:

  • Each column is a feature describing something about the each cell (=row). So from left to right it’s quite straightforward. You have the name, the classification of the object, its X and Y locations (n.b. origin is at the top left of the image), the area and perimeter of the nucleus, its circularity (e.g. a perfect circle would have a circularity of 1), and its Feret diameter (used here and some definition here). If you want more information Google is probably the best option.
  • Now if you use any of these features for a classification task, it basically means that the classifier will use each cell detection (that has a class assigned) as a training object and will use its feature values to create a model (e.g. Random Tree, ANN, …) and make the best predictions for other detections.

Please refer to this for info about classifying your cell detections. Bearing in mind that the classifiers are being updated and improved for the next releases :slight_smile:

1 Like

Thank you very much. I got it. Could you please tell me why we need to color transform the add intensity features?Under what circumstances should we select the following corresponding options?The other one is what does DAB mean by color deconvolution?
image

I would recommend searching for the forums about some of the terms like color deconvolution, as an understanding of what is happening there would be very important for any brightfield image analysis. There is plenty of information here.

Also, a useful website on the topic.
https://blog.bham.ac.uk/intellimic/g-landini-software/colour-deconvolution/

Note that all of those options can be seen in the Brightness/Contrast window.
image
So that you can decide which are useful given your circumstances and experiment. It will vary widely from person to person, but usually DAB is the most important since… that is usually the experimental condition.

1 Like

I have learned a lot from wikipedia. Thank you for your help!
Sorry, there is a small problem. What does nucleus shape: area um^2 mean? I got the result by adding shape features, but I don’t know what it means.

That should be almost the same as the Nucleus: Area at the top of the measurement list. I suspect any slight difference would be due to whether you have “Smoothing” checked when creating the cells, which would change this second measurement slightly.

Essentially. It is the area… of the nucleus.

Thank you for your answer!Have benefited a lot .

1 Like

Sorry, there’s a new problem again, how is the measurement of the Add coherance grain feature such as Hematoxylin (25um) coherance calculated?

Coherence is based on the structure tensor - see wikipedia. For the details of QuPath’s implementation, it’s best to rely on the code.

Be sure to check out Measure → Show measurement maps. I think that the easiest way to get a feeling for what specific features mean by visualizing them. If you aren’t familiar with this command, it is shown in the second YouTube tutorials (Video #19).

1 Like