Wavelet features with "flat region"?

Hi everyone!

I am doing a feature extraction and one of the problems I am having is some “flat regions” in the wavelet features of some events (not all dataset and not always the same features), like: wavelet-LHH, “glszm”, “SizeZoneNonUniformity”, “1.0”
wavelet-HLH, “glszm”, “GrayLevelNonUniformity”, “1.0”
wavelet-HLH, “glszm”, “SizeZoneNonUniformity”, “1.0”
wavelet-HHH, “glszm”, “SizeZoneNonUniformity”, “1.0”

My interpolator is ‘sitkBSpline’, with a resampledPixelSpacing of [2, 2, 2], as I have slices from 2.5 to 8 mm. I selected my binWidth based on the range feature of the whole dataset and found that the best value is 3 (it is within the ideal range 16-128).

For some events, the above problem disappears when I set resampling to 2.5,2.5,2.5 but for others, it does not work. I have also tried changing the binWidth.

Can you tell me if the values of these parameters (binWidth and resampledPixelSpacing) are ok, and if this “flat regions” are normal? Or could this also be a problem with the wavelet (I have not entered any parameters into this filter)?

Thank you for your attention!

Wavelet filtering can result in quite different values. You can review the range feature extracted from the filter to be sure. If necessary, you can set a different binWidth for the wavelet-filtered images, though it is not possible to do this for e.g. just the HLH filtered.

1 Like