Choosing training dataset

I have hundreds of images and Í would like to pick 10 representative cases (covering highest and lowest contrast) to customize the default ‘positive cell detection’ setting. For this purpose, I am running intensity measurement on color transforms for 1. optical density, 2. Hematoxylin (deconvolved) and 3. DAB (deconv). Can anyone comment on it if this is the right way to do?
Many thanks.

You may have better luck performing the cell detection, then looking at the distribution of “measurement of interest” across the data set. You will rarely get a nice bimodal distribution, but it might give you an idea of the data distribution when you look at it in Excel. In M12 you can easily select a few measurements of interest to export for the whole project (Measure menu).