How to apply spectral flatness measure on the signal/spectrogram using Matlab?

I am using MATLAB. I found a couple of spectral flatness measure functions written for MATLAB. My question from here is how to actually use this measure and apply it to either the time-domain signal (vector) or a spectrogram (matrix) in order to separate out the noise from the non-noise signals? In other words, how to use the spectral flatness measure to actually do the filtering?

Basically, how to go from a feature, such as: https://www.mathworks.com/matlabcentral/fileexchange/51848-spectral-flatness-a-

OR

http://www.audiocontentanalysis.org/code/audio-features/spectral-flatness/

and then do the following with it: https://www.thinkmind.org/index.php?view=article&articleid=signal_2016_3_10_80021

OR

http://www.ient.rwth-aachen.de/services/bib2web/pdf/BeRo13.pdf

You do not have to explain to me using this specific feature, you can explain using any feature vector(or matrix) that you know and that was then used to do something. For example, some features allow you to track polyphonic pitch and use that in order to separate the main melody from the mixture signal. I am asking how I would use spectral flatness to separate mixture into noise and non-noise components. I hope it's clear now.