Clarifying cluster versus subcellular spot

I am trying to run some quantifications of subcellular spot detection from exported annotations from QuPath for in situ hybridization data that I ran both cell segmentation and spot detection on with cluster splitting. For each regions I have three measures:

Num Detections
Num Subcellular cluster: DAB object
Num Subcellular spot: DAB object

My questions are:

Does num detection = # cells + # clusters + # spots?
Does num subcellular spots = # detected spots + # spots that come from split clusters?

Or do I need to sum the # clusters and # spots to get the total number of spots detected?

Sorry if these are basic! Thanks!

I would say there are two main things to be aware of when using subcellular detections for counts.

  1. If you sum up the areas of subcellular detections, you will usually get a lower number than the estimated spot count because spots below the size of a single spot will still count as 1 spot.
  2. Each spot or spot cluster is an object. The estimated number of spots in a cluster is the area of the cluster divided by the estimated spot size you chose.

If you chose 1 square micron for your spot size, the estimated spot count for a subcellular cluster will equal the area of that cluster. The estimated spot size of a spot that is smaller than one micron will still be one.
Example assuming the estimated spot size is 1.0.

One cell has two spots and one cluster.
Spot 1 is 0.5 um^2
Spot 2 is 1.0 um^2
Subcellular cluster 1 is 2.5 um^2

There are 4 detections in this example, one cell and it’s three children.
The total area if you summed the areas of those objects would be 4um^2.
In the cell’s measurement list:
The estimated spot count (“Subcellular: DAB: Num spots estimated” ) would be 1+1+2.5 = 4.5 estimated spots.
Num single spots would be 2
Num clusters should be 1, but upon testing it does not seem to be registering clusters for me. That might have been lost after the change to the object type for subcellular objects.

In the annotation measurement list, you would get
4 detections
1 cluster
2 spots
You cannot tell the number of total estimated spots from the annotation measurement list, since you do not know what each cluster represents (it could be 2.5 spots, it could be 50).

Thanks that is helpful! So the ‘Num Subcellular spot: DAB object’ does not already include the number of dots that was estimated by using the cluster area? For reference I selected ‘include clusters’ and gave an estimated spot size.

1 Like

If it is specifically spots, it will only include spots, not clusters, correct. The total estimate is the Num spots estimated.

Ah I see now that I can export those measures if I select ‘Detections’ and not ‘Annotations.’ Thanks so much for clarifying!

Yes, you can also apply means of measurements per class to parent annotations if you want to dive into the scripting, but if you can handle that all outside of QuPath, that might be best. I mostly bother with scripting summary results so that I can immediately see the effects of settings changes.

Good tip. I am doing most my quantification in R. Looks like the measurements I want to export are per cell where if I am interpreting it correctly the Num.estimated.spots includes both single dots along with the inferred spot number from cluster area.

1 Like

Haha, having “fun” trying to remember how to code in R right now to process some Imaris exports!

1 Like