Qupath: Fluorescent image analysis using colocalization of objects in different channels

I am using the latest milestone 6 for my image analysis.
I am trying to analyze fluorescent images taken on delta vision systems. I am staining for 3 surface markers and DAPI stain. Below are the jpeg images of representative sample image.

First I tried to segment the DAPI channel for getting the cells by expanding the nucleus detection. Then using the intensity measurements within the cells to define positive cells in other channels. Since all are surface markers in my images, so I don’t get exact detections.
In order to overcome this I am trying to segment all the channels separately and want to do colocalization to get single, double or triple positives. But in qupath, I can segment only one channel and as I segment the second channel, I loose the segmentation for first channel.
I will appreciate any suggestions in this regard or even any other suggestion for this kind of image analysis.

I would recommend looking into these options, thought the code may be written for different versions of qupath at this point. The easiest way would be to use a large enough cytoplasmic expansion when defining the cells, and then subcellular detections.

Possibly useful, though not exactly for what you are describing.

Thanks for your suggestions. I am working on sub cellular detections and looks like it will work.

  1. I am detecting my nuclei in DAPI channel with “Cell Detection”.
  2. I am using “Sub cellular detection” for other three channels (Cy5, TRITC, FITC).
  3. Now I want to build a classifier based on different parameters but I am unable to locate certain options for building the classifier, similar to one shown in this post (Guide: QuPath Multiplex analysis workflow (detailed))

Thanks again.

  1. Could you be more specific? Does the script not run or what does it look like on your system? I don’t know what certain options are without more information.

I want to build a classifier.
Classify>Object Classification>Create Detection Classifier

Is it the right option to use for the classifier (remember I am using milestone 6).
Thanks for your quick reply.

That isn’t the same classifier. The link above the image you copied points you to the script here:

What you are looking at is the trained detection classifier. It doesn’t use thresholds, it uses training sets. More description of that in the top post of the link in this post.