MeasureImageQuality

Hello,

I am using the MeasureImageQuality module to track the image quality over my assay. I would like to verify the meaning of the “MeasureImageQuality_Threshold” metric (it does not seem to be explained in the documentation). Would it correspond to some kind of global segmentation threshold computed for the image? (if so, what would be the settings for this segmentation threshold)?

Thank you,

Ernest

Hi Ernest,

MeasureImageQuality in fact uses the same thresholding methods as the Identify modules (that is, the choices are the same- so you can pick Otsu, MoG, etc just as you can in IdentifyPrimary). This measurement is meant to give you a sense of what the range of thresholds would be for an image set. The settings are exactly the same as for IdentifyPrimary, so reading the help for that module should give you an idea of what the different algorithms do.

How this can be useful: If you run an ‘image quality’ pipeline before you do any segmentation, this can let you know what the upper and lower bounds should be for your images in IdentifyPrimary or IdentifySecondary. Very bright or very dim images usually cause algorithms (especially Otsu) to fail badly, often causing your pipeline to take an unneccesarily long amount of time, as Otsu finds thousands of “cells” (really just the algorithm failing and segmenting out of control).

Another way it can be useful is if you simply to decide to take the mode of histogram of thresholds for the whole image set and use this as a hard threshold- again, helpful if you’re worried about an algorithm failing. The suggested threshold will also tell you which images are very bright and which are very dim, and this can sometimes be loosely translated into ‘blurry’ or ‘debris-containing’- which you may want to exclude from analysis altogether. Focus score is really supposed to be a more direct measure of focus, but we’ve found that threshold sometimes correlates as well.

~kate