new feature QuPath 2.0 - pixel classification

[Filip Bochner]

Hi,

I'm trying to apply the new feature to segment the areas positive for Sirius Red staining. The goal is to get the fraction area (Sirius Red area divided by the tissue area). I was trying to follow the tutorial:

https://qupath.github.io/QuPath-v0.2.0

but after pressing "show output classification", I get a message that No classifier has been trained yet. I marked the collagen areas with wand tool, as it was presented in the tutorial, and assigned a color to them. What did I miss?

Thanks!

Filip

[Pete]

It sounds like you need at least annotations with at least one other classification.

You can use the special 'Ignore' classification for background/artifacts that you don't want to measure, and whatever other class you like for tissue (except 'region' which is also special...).

More documentation on the pixel classifier to come soon...ish.  However I'd suggest it's worth experimenting with annotating using the polyline tool rather than the wand, since this will give fewer samples.  In some cases I find the polyline tool helps me train a better classifier, faster - but in other cases I find the brush or wand to be more effective.  Either way, the pie chart should help give a visual guide to how many training samples there are for each class.

[micros...@gmail.com]

Do you have the minimum of two classes required to train a classifier?

[micros...@gmail.com]

Ah, I think I just did what you did! I suspect you want to click "Live Prediction" option rather than Show next to Output. Show will just create a small window around the area you have selected (it seems).