pyradiomics w/ fixed bin width: bin number varies significantly across derived images

Hi all,

When using the fixed bin width setting, the median number of bins per patient varies a lot across the different derived images and the original image. I know this can be in part addressed by selecting a different bin width for e.g. original, LoG filtered and wavelet images. However, the same issue occurs across different e.g. wavelet decompositions which only differ in their directionality. Here’s an example:

Image → median number of bins across patient cohort
wavelet-LLH → 172.5
wavelet-LHL → 210.7
wavelet-LHH → 35.0
wavelet-HLL → 197.6
wavelet-HLH → 33.6
wavelet-HHL → 35.2
wavelet-HHH → 8.8
wavelet-LLL → 1132.2

Anyone familiar with this behaviour? How can this be addressed? Can I specify different binWidth settings for the above decompositions?

Any help would be appreciated :slight_smile: Thanks much!