Reproducibility of Radiomics Results with Pyradiomics and Parenchymal Analysis

Community Radiomics
1

Hi there,

I am trying to see if the Radiomics module analysis for the lung segmentation I performed using CIP is consistent with that derived from Parenchymal Analysis. This is partly because the Parenchymal Analysis is no longer available on the later versions of 3D Slicer. Thus far, most of the parameters are the same but I keep getting different values for kurtosis. Just wondering what went wrong?

Here is an example

segmented example
segmented example775×673 77.4 KB

radiomics sample
radiomics sample1462×1308 334 KB

parenchymal analysis sample
parenchymal analysis sample1755×452 84.7 KB

As you can see, all of the matching parameters have same values except for kurtosis. Perhaps the formulas used for the modules are different?

2

This is a known confusion. What happens is that parenchymal analysis gives you the excess kurtosis, while pyradiomics gives you the true kurtosis. It’s a static difference of 3 (derived from the normal distribution, where kurtosis = 3, and excess kurtosis = 0).