neurolabusc / dcm_qa

DICOM to NIfTI/BIDS validation script
BSD 2-Clause "Simplified" License
11 stars 2 forks source link

Get some multi-echo samples #8

Closed yarikoptic closed 5 years ago

yarikoptic commented 6 years ago

E.g. here is one (non functional, so somewhat not that interesting): http://datasets.datalad.org/?dir=/dicoms/dartmouth-phantoms/bids_test6-PD+T2w/

ATM there seems to be none:

$> git show-ref HEAD
c7bab67f14105031776e194d47961749f264e209 refs/remotes/origin/HEAD
$> git grep EchoNumber

and it would be great to test/make sure that dcm2niix could work consistently across different manufacturers (ref issue in heudiconv: https://github.com/nipy/heudiconv/issues/162#issuecomment-385412124)

I could provide more samples from Siemens but unfortunately we do not have any EPI multi-echo sequences I believe :-/

neurolabusc commented 6 years ago

@yarikoptic - the current dcm_qa is designed to simulate problematic datasets or rich details. Since we test every build of dcm2niix (separately for Windows, MacOS, Linux), I have tried to keep the disk requirements small.

I can see the value of an expanded testing dataset, but I do not think that Github is the correct platform for this. I have not had time to investigate, but DataLad or GitLab may be better ways to host larger datasets.

I would suggest you fork dcm_qa - either on Github or another platform. I would be happy switch dcm2niix to test your more extensive battery once it is validated.

The dcm2niix wiki includes diffusion examples from GE, Siemens and Philips. Hopefully, we will have some new modern Philips DTI scans soon.

For images, I would prefer using a human head over phantoms: these allow you to infer image rotations and bvec calculations. I would also suggest placing a fiducial marker on one hemisphere (you can see a saline bag on the left temple in several of the linked datasets).

neurolabusc commented 6 years ago

@yarikoptic - I acquired 3 different forms of [https://www.nitrc.org/plugins/mwiki/index.php/dcm2nii:MainPage#Multi-Echo_MRI multi-echo MRI] for validation. First, field maps are the Siemens Product and are saved as 2D slices. Second, the CMRR research EPI T2* fMRI/resting state sequences saved as mosaics. Finally, the MGH MEMPRAGE T1 scans. The samples proved useful already - I already made changes to dcm2niix. I do think these require too much disk space for dcm_qa, but would be useful for a future testing repository.

yarikoptic commented 6 years ago

Great, thanks! If you are interested to go git-annex (what datalad uses) route for the ultimate collection, I would be happy to help to establish it. git-annex on Windows is a bit of a pain but for this minimal purpose of his getting the content, it should suffice. We could host content somewhere or even just use https://web.gin.g-node.org which does provide some storage and supports git annex... Main clone of the repository could still reside on GitHub, just content live elsewhere .

neurolabusc commented 6 years ago

I honestly have too much on my plate to curate this at the moment. If you want to step up to the challenge, you can merge dcm_qa with the files at the dcm2niix wiki. I am sure many will find it useful.