Binwidth Optimization

Hello everyone,

I’m trying to extract the radiomic features from MR images using Pyradiomics extension on 3DSlicer.

I know that the default setting for the binwidth is 25. However, I am not sure which value is more suitable for my studies. How can I optimize this value? Some features cannot be calculated when I use the default value.

Thank you for your help.

  1. Extract firstorder features for your dataset, with all preprocessing you intend to use.
  2. Look at the value for firstorder - Range; choose a binwidth so, that for the majority of your scans, the binwidth will result in ~10-100 bins. So if your range is, saty [50 - 500], use binwidth 5.

Thank you for your kind answer.

I want to ask again, just to be sure.

In my FO Range, my minimum number is 1.46 and my maximum number is 7.58. What’s your suggestion for binwidth value?

In that case, I would advise a binWidth of about 0.1, this will yield 15 bins for your minimum and 76 for your maximum, both of which are in the range of [10, 100] bins.

What if we have a range of [50 12000] or [10 1000]? We can garantee that both the min and max stand on the [10 100] interval. What should we do in those cases?