Many 0 and 1 appear in the extracted features

Hi, I used Radiomics module to extract PET image features. I set resampled voxel size to 1,1,1 and bin width to 64 based on the literatures , and the result shows that there are many 0 and 1 in GLCM, GLDM, GLRLM, GLSZM and NGTDM. I don’t know where the settings are wrong, can you help me check? Thank you!


Hello @lrc_09,

What you are seeing are typical results when extracting from a ‘flat region’. See the online PyRadiomics documentation for more details. TLDR: The binwidth parameter is set too large (default 25, which is too much for SUV values).

The ‘bin width’ of 64 from literature is incorrect, the 64 you encounter in literature is 64 bincount. In PyRadiomics we advise against using a fixed binCount, details for this are also found in the PyRadiomics documentation.