Auto3DSeg and transforms

[adrianvallsc]

I am running Auto3DSeg on a public dataset. The main problem is dimensions of labels and images do not fit and are not homogeneous.

With standard MONAI pipelines I was using transforms such as ResampletoMatch to make them homogeneous. When I run Auto3DSeg as I could not find how to customize the transforms it fails:

RuntimeError: Caught RuntimeError in DataLoader worker process 0.
Original Traceback (most recent call last):
File "/usr/local/lib/python3.10/dist-packages/monai/transforms/transform.py", line 141, in apply_transform
return _apply_transform(transform, data, unpack_items, lazy, overrides, log_stats)
File "/usr/local/lib/python3.10/dist-packages/monai/transforms/transform.py", line 98, in _apply_transform
return transform(data, lazy=lazy) if isinstance(transform, LazyTrait) else transform(data)
File "/usr/local/lib/python3.10/dist-packages/monai/apps/auto3dseg/transforms.py", line 82, in call
raise ValueError(
ValueError: The label shape torch.Size([112, 112, 72]) is different from the source shape torch.Size([352, 352, 29]) /content/drive/MyDrive/segmentation/LABELS/sub-strokecase0242_ses-0001_msk.nii.gz.

[KumoLiu]

Hi @adrianvallsc, from the error message, I think you have a different size between the image and the label. You may need to check this data "/content/drive/MyDrive/segmentation/LABELS/sub-strokecase0242_ses-0001_msk.nii.gz".
Thanks!

[adrianvallsc]

Yes, I guessed this was the reason. My question was wether it was possible to apply custom transforms to Auto3DSeg run

[KumoLiu]

Hi @adrianvallsc, I think the error was before training, it was in the data_analyzer step, so adding your custom transforms may not solve this error. You may need to check the data first. :)

MONAI/monai/apps/auto3dseg/data_analyzer.py

Line 299 in 575954c

EnsureSameShaped(

Thanks!