Higher order radiomics features - cerr/CERR GitHub Wiki

The higher order features can be calculated in 2 and 3 dimensional setting. The 2-d calculation might be preferred if the in-plane resolution differs a lot from the slice thickness. For the 2-d calculation, only 2-d neighborhood and/or directionality is considered; whereas 3-d neighborhood and/or directionality is used for the 3-d calculation of higher order matrices (for example GLCM, RLM, NGTDM etc.). These matrices are then reduced into scalar features. This corresponds to the workflow depicted in figures 4.2 (d) and (e) in the IBSI document. Note that the Gray Level Co-occurrence Matrix (GLCM) and Gray Level Run Length Matrix (GLRLM) based features can be computed in two ways: (i) computing the scalar feature/s separately for each directional offset and then taking the mean/max/std. etc. (ii) adding contributions from all the directional offsets into a GLCM/GLRLM and then computing the scalar feature/s.

• Gray Level Co-occurrence Matrix (GLCM)
• Gray Level Run Length Matrix (GLRLM)
• Gray Level Size Zone Matrix (GLSZM)
• Neighborhood Gray Tone Difference Matrix (NGTDM)
• Neighborhood Gray Level Dependence Matrix (NGLDM)