How to quantify small vesicles in IF images


I am trying to quantify small vesicles in IF images (see attached) which are very tiny dots in the red channel. The cells in this image are fibroblasts, so using the cell detection won’t work, as the detection won’t recognise the full length of the cell. In the image I also have stain in the green channel for the cytoskeleton, which gives me the shape of the cell.
I’ve tried using the pixel classifier and the object classifier, but neither gave me accurate results. Maybe I’m not using the correct settings? Or is there a different way I can use QuPath for this?

Thanks a lot!

Have you tried filtering individual channels for your classification in the pixel classifier options? (Feature > Edit > Channels). If the spots are very small you can also higher the resolution of the classification, to be sure not to miss small dots.

I will give it a try to:

Or maybe you can try to use the approach I used in this paper:

In the Supplementary material at Quantitative imaging analysis of CD147 internalization
you can find the description of the algorithm I implemented in Fiji.

If you are interested I can search on my computer where I put that plugin and we can try if it works on your images.

Unfortunately, I’ve quite 0 experience with quPath…

Emanuele Martini

For the pixel classifier, you are using the Simple thresholder/Create thresholder, yes? If so, and at the highest resolution setting as @melvingelbard said, that is probably the best option you have within QuPath. Though, a thresholder is going to be more dependent on your samples being taken in a similar way with similar settings (versus the pixel classifier which might detect edges).

If that is what is not working, can you explain in more detail how it is not working? Not giving accurate results does not tell us what is going wrong. If you want to quantify “per cell” you could either divide by the number of nuclei (assuming the image is bigger than this and the whole cell fits in the image) or else you would need some other marker to determine the cell borders.

Thanks all!
I’ve tried what @melvingelbard suggested and it worked!
@Research_Associate what I meant by not getting accurate results was that the count I was getting was completely off. For example, for the image I’ve added the program was giving me 54 red dots when you can see that there are clearly more than that. But filtering the channel definitely helped!