Intensity normalization question

[dscarmo]

Hello,

I have noticed in your code you use ImageNet mean-std normalization, which is for images in the 0-1 range.

Therefore, do you use min-max normalization to put Hounsfield Unit input images (original CT values) into the 0-1 range? Did you do that per volume? I haven't found details about that in the paper.

I want to make sure I am reproducing your work correctly, since the network performance can be affected greatly by wrong normalization.

Thank you

[pobeidy1]

Greetings, Nice code, thank you! I have got the same question