Question on ADNI

[EmreTaha]

Hi, thanks for the nice work and detailed pipeline. I have couple of questions on the ADNI, It would be much appreciated if you can clarify.
I created a csv file for ADNI phases based on your flow chart. For T1w scans from ADNI, did you use any scan other than mprage? Also apparently, ADNI doesnt support downloading nifti version of the raw scans, only 2D cross sectional dicoms (https://groups.google.com/g/adni-data/c/Rs9NE8R4mDg), which tool did you use for converting dicom to nifti (I assume a conversion is applied since the preprocessing requires nifti)?
Thanks in advance!

[LemuelPuglisi]

Hi!

Regarding ADNI, I’ve only used MPRAGE images.

About the conversion to nifti, you can use dcm2niix: https://github.com/rordenlab/dcm2niix

Best regards,
Lemuel

[JornyWan]

hi, I have processed ADNI dataset and tried to train the LDM based on pretrained Autoencoder, but the training got loss=nan and the generated images from image_to_tensorboard are noise while I didn't make any change of the [train_diffusion_unet.py] as well as other files. I'd really appreciate it if this could be solved, thanks for your great work!

[JornyWan]

FYI, i added the weights_only to False in torch because the original training code returns error as the following:

Traceback (most recent call last):
File "/BrLP/scripts/training/train_diffusion_unet.py", line 152, in
z = trainset[0]['latent']
File "/.conda/envs/brlp/lib/python3.9/site-packages/monai/data/dataset.py", line 112, in getitem
return self._transform(index)
File "/.conda/envs/brlp/lib/python3.9/site-packages/monai/data/dataset.py", line 424, in _transform
pre_random_item = self._cachecheck(self.data[index])
File "/.conda/envs/brlp/lib/python3.9/site-packages/monai/data/dataset.py", line 386, in _cachecheck
return torch.load(hashfile)
File "/.conda/envs/brlp/lib/python3.9/site-packages/torch/serialization.py", line 1470, in load
raise pickle.UnpicklingError(_get_wo_message(str(e))) from None
_pickle.UnpicklingError: Weights only load failed. This file can still be loaded, to do so you have two options, �[1mdo those steps only if you trust the source of the checkpoint�[0m.
(1) In PyTorch 2.6, we changed the default value of the weights_only argument in torch.load from False to True. Re-running torch.load with weights_only set to False will likely succeed, but it can result in arbitrary code execution. Do it only if you got the file from a trusted source.
(2) Alternatively, to load with weights_only=True please check the recommended steps in the following error message.
WeightsUnpickler error: Unsupported global: GLOBAL monai.data.meta_tensor.MetaTensor was not an allowed global by default. Please use torch.serialization.add_safe_globals([MetaTensor]) or the torch.serialization.safe_globals([MetaTensor]) context manager to allowlist this global if you trust this class/function.