nnUNet output not aligned properly with the original scan

[Flock1]

Hi,

I trained nnUNet on my own data and I was wondering if you could help me alignment issues.

This is the command I use to run predictions:

nnUNetv2_predict -i /path_to_nnUNet/nnUNet/nnUNet_results/Dataset001_SpinalCordSegmentation/imagesTs -o /path_to_nnUNet/nnUNet/nnUNet_results/Dataset001_SpinalCordSegmentation/output -d 001 -c 3d_fullres -f 4 --verbose
And the outputs I get, even though the segmentations look good, are not aligned with the original segmentation file. Hence, I am unable to calculate the dice score. Can someone help me?

[Flock1]

@saikat-roy, can you help me with this issue?

[Flock1]

An update, I realized that for the majority of scans, the output aligns well with the original scan. It's only for some scans it's not working. So I was wondering if that's the actual output of nnUNet or if the affine transformation is wrong. Do you have any suggestions on how to confirm this?