I noticed some Issues with missing features from the IBSI recommendation in Pyradiomics.
I went through the feature definitions and compared them (IBSI: https://pubs.rsna.org/doi/full/10.1148/radiol.2020191145, Pyradiomics: https://pyradiomics.readthedocs.io/).
Some features like Compactness 1 & 2, Asphericity, Sum variance, Dissimilarity, etc. are equal or highly correlated to other features. I think it makes sense to exclude those.
But I detected a bunch of missing features that are included in the IBSI feature definition but not in Pyradiomics.
I was wondering if there is any justification I am missing or if it is possible to include those into the feature set?:
Intensity-based statistical features:
Intensity-based median absolute deviation
Intensity-based coefficient of variation
Intensity-based quartile coefficient of dispersion
Centre of mass shift
Volume density (axis-aligned bounding box)
Volume density (oriented minimum bounding box)
Area density (oriented minimum bounding box)
Volume density (approximate enclosing ellipsoid)
Area density (approximate enclosing ellipsoid)
Volume density (minimum volume enclosing ellipsoid)
Area density (minimum volume enclosing ellipsoid)
Volume density (convex hull)
Area density (convex hull)
Geary’s C measure
Local intensity features
Local intensity peak
Global intensity peak
Intensity histogram features
Intensity histogram mode
Intensity histogram mean absolute deviation
Intensity histogram robust mean absolute deviation
Intensity histogram median absolute deviation
Intensity histogram coefficient of variation
Intensity histogram quartile coefficient of dispersion
Minimum histogram gradient
Minimum histogram gradient intensity
Intensity-volume histogram features
Volume at intensity fraction
Intensity at volume fraction
Volume fraction difference between intensity fractions
Area under the IVH curve
Hope for help.
Thank you in advance.