nnUNet icon indicating copy to clipboard operation
nnUNet copied to clipboard

Final Contouring

Open tedi14 opened this issue 1 year ago • 3 comments

Hello,

I have been able to train a model on my own data. Now I want it to produce contours on an unseen CT data and want to be able to see it. Can anyone direct me to be able to do that? I am sorry if its a simple thing I am just not sure what to do next.

Thanks!

tedi14 avatar Aug 13 '24 18:08 tedi14

Hey @tedi14. How about using np.gradient to calculate x, y and z gradients and adding them, of binarized output maps per class. This should give you contours of the segmented structures.

saikat-roy avatar Aug 13 '24 19:08 saikat-roy

Hi @saikat-roy, I was hoping to run a few predictions on a few patient ct data sets and then take those predictions as nii.gz files and then convert them to dicom with their rstruct contours to examine them and see how the segmentations look. Is there a way to code this or run the first few steps getting a prediction segmentation of a new patient. What are the steps in running this?

Thanks!

tedi14 avatar Aug 14 '24 09:08 tedi14

Hi @tedi14,

Converting the nifti segmentations to an RT struct is not that straightforward, you can use PlatiPy, you can also try our tool, nifti2dicom, but we convert it to DICOM seg objects.

Cheers, Lalith

LalithShiyam avatar Aug 14 '24 17:08 LalithShiyam