This parameter file (below) works fine and calculates about 2200 features but I am still concerned that any of the settings may contradict each other, or be in the wrong order and therefore deliver incorrect values.
If the default normalisation (z-score) narrows the pixel intensity range to -3/+3 this should mean that binWidth cannot be lower than 6? And that is for only one bin?
Is it then better to set the binCount?
Would normalizeScale be useful to define, say to 100?
What is the optimal binCount for MRI?
Is any more detail needed for the image types or feature classes?
Feedback on this file would be of tremendous help to me!
Cheers, Marko
setting:
normalize: true
resampledPixelSpacing: [1, 1, 1]
interpolator: 'sitkBSpline'
binCount: 32
label: 1
imageType:
Original: {}
LoG: {'sigma' : [1.0, 2.0, 3.0, 4.0, 5.0]}
Wavelet: {}
Square: {}
SquareRoot: {}
Logarithm: {}
Exponential: {}
Gradient: {}
LBP2D: {}
LBP3D: {}
featureClass:
glcm:
firstorder:
shape2D:
shape:
glrlm:
glszm:
gldm:
ngtdm: