Problem on prepocessing

[mikami520]

Hi all, I started doing the registration on multiple segmentations (.seg.nrrd) and save it as (nii.gz). Then I used the nnUnet to preprocess the dataset. It gave me errors when cropping the images. As I checked my segmentations and training CTs, they are all 3D images. I have no idea about this issue. Here is the screenshot, could anyone help me with this issue?

[duytq99]

Hello @mikami520 Can you please post the command you've used?

[mikami520]

Hello @mikami520 Can you please post the command you've used?

Hi @duytq99 Thanks for your response. I use "nnUNet_plan_and_preprocess -t 004 --verify_dataset_integrity" for preprocessing dataset, where 004 is the task id

[duytq99]

Hi, Could you insert the line print(data.shape) to inspect data dimension at https://github.com/MIC-DKFZ/nnUNet/blob/a7d1d875e8fc3f4e93ca7b51b1ba206711844d92/nnunet/preprocessing/cropping.py#L94 and provide the output? Thank you.

[mikami520]

Hi, Could you insert the line print(data.shape) to inspect data dimension at

https://github.com/MIC-DKFZ/nnUNet/blob/a7d1d875e8fc3f4e93ca7b51b1ba206711844d92/nnunet/preprocessing/cropping.py#L94

and provide the output? Thank you.

When I insert that line, do I need to re-install the nnUnet library?

[duytq99]

Yes. You need to do !pip install -e . otherwise you can run the script directly using python nnunet/experiment_planning/nnUNet_plan_and_preprocess.py -t 004 --verify_dataset_integrity That data.shape should have 4 elements. If not, I guess there something wrong and you need to trace back more. I usually inspect the value or shape of variable. Hope this can help you figure out what's wrong.

[TaWald]

Due to the thread being stale and the release of nnUNetV2, I'll close this issue for now. Should the issue still remain @mikami520, or someone else run into a similar issues, feel free to create a new issue or reopen this one.

Also thanks a lot @duytq99 for the support :)