search

Parietal-INRIA / DiFuMo

Dictionary of Functional Modes for brain imaging
https://parietal-inria.github.io/DiFuMo
23 stars 5 forks source link

Error when downloading atlases #16

Closed nikitas-theo closed 3 years ago

nikitas-theo commented 3 years ago

It appears there is some problem in nilearn.utils._fetch_files . I include the error message, files = _fetch_files(data_dir, files, verbose=2) File "/home/semeval/anaconda3/lib/python3.7/site-packages/nilearn/datasets/utils.py", line 721, in _fetch_files verbose=verbose, session=session) File "/home/semeval/anaconda3/lib/python3.7/site-packages/nilearn/datasets/utils.py", line 797, in _fetch_files raise IOError('Fetching aborted: ' + abort) OSError: Fetching aborted: Dataset has been downloaded but requested file was not provided: URL: https://osf.io/3vrct/download Target file: /home/semeval/nilearn_data/difumo_atlases/256/maps.nii.gz Downloaded: /home/semeval/nilearn_data/difumo_atlases/d7aa641714f90074ed0175b5ebcb43b8/download

KamalakerDadi commented 3 years ago

Hi, Thank you for raising an issue. It is fixed now. Please let me know if issue still exists.

KamalakerDadi commented 3 years ago

Sorry for the noise.

nikitas-theo commented 3 years ago

Thank you for fixing it ASAP. Can i quickly ask another question, although it is not issue related. Resampling maps in a dataset of shape=(53, 63, 52, 857) seems to take forever. Is this computationally intractable or is there a problem with my system?

bthirion commented 3 years ago

It depends on whether the transformation is a pure translation (in which case it should be fast) or not (-> trilinear interpolation is then used, which is more costly).

nikitas-theo commented 3 years ago

Thank you for the reply, I am sorry it is actually the [NiftiMapsMasker.transform_single_imgs] Extracting region signals step that seems to be stuck

KamalakerDadi commented 3 years ago

it is actually the [NiftiMapsMasker.transform_single_imgs] Extracting region signals step that seems to be stuck

I assume you are using default resampling strategy i.e., resampling functional data to maps. Is this the step that is stuck?

nikitas-theo commented 3 years ago

I am using default settings. I am posting the terminal output, the step "Extracting region signals" never finishes. 

[NiftiMapsMasker.fit_transform] loading regions from /home/semeval/nilearn_data/difumo_atlases/1024/2mm/maps.nii.gz
Resampling maps
[NiftiMapsMasker.transform_single_imgs] Loading data from Nifti1Image(
shape=(53, 63, 52, 857),
affine=array([[  -3.,    0.,    0.,   78.],
       [   0.,    3.,    0., -112.],
       [   0.,    0.,    3.,  -70.],
       [   0.,    0.,    0.,    1.]])
)
[NiftiMapsMasker.transform_single_imgs] Extracting region signals
bthirion commented 3 years ago

@KamalakerDadi IIRC there is a 3mm-version of the Difumo atlas. using it instead does bypass the resampling step. Am I right ?

KamalakerDadi commented 3 years ago

@KamalakerDadi IIRC there is a 3mm-version of the Difumo atlas. using it instead does bypass the resampling step. Am I right?

I don't think so unless resampling_target=None. We can resample mask, functional data to maps resolution using resample_img before using NiftiMapsMasker, then simply putting resampling_target=None in NiftiMapsMasker may avoid this issue.

That's one way or try maps resolution to 3mm.

KamalakerDadi commented 3 years ago

@nikitas-theo i assume its ok