[Mrtrix-discussion] Brain Imaging Data Structure - call for feedback

[Chris Filo Gorgolewski]

Dear neuroimaging community,
Most of our daily work involves data - acquiring, organizing, cleaning,
analyzing, understanding, and explaining data. MRI scanners produce a lot
of complicated outputs with rich metadata needed for further analysis. In
addition all of the external measures of behaviour and individual
differences contribute to the complexity of acquired data. It is easy to
get lost in this chaos especially for early scientists new to the field.

How we store, organize, and describe neuropsychological data has been an
individual issue. Each researcher had their own way of describing data.
Such approach can be problematic when one dataset needs to be used by more
than one person. Imagine a situation when as a PI you want your postdoc to
reanalyze an old dataset acquired by a long gone PhD student just to
discover that bits are missing and the rest is an unreadable Excel
spreadsheet.

We want to solve this problem. For the past couple of months we have been
working on a specification describing how to organize and describe
neuropsychological data obtained from fMRI (task and rest), structural, and
diffusion experiments. Our goal was to make it easy to adopt and mirroring
practices people already follow, but at the same time friendly for
developers. The new standard is called Brain Imaging Data Structure (BIDS -
http://bids.neuroimaging.io/) and it's based on folders, NIFTI, JSON and
tabular files. It does not require any software (such as a database). It's
intended to capture raw data, but it's easy to extend with derivatives.
This is how a BIDS dataset looks like:
https://55b6842373553c2422bd99ac533b084533ea6cb7.googledrive.com/host/0B2JWN60ZLkgkQm9ZazZCWS1VNVE/ds114/
Here's the full spec: http://bids.neuroimaging.io/bids_spec0.4.pdf

Having a common formalized way of organizing and describing data have
advantages beyond sharing across members of the same lab. Workflow
developers can build tools that will be able to preprocess your data with
very little user input. It will be also easier import such formatted
dataset into databases (XNAT, COINS, Scitran, NiDB, openfmri etc.) and
share it with wider public. We already started writing tools (and
OpenfMRI2BIDS converter: https://github.com/INCF/openfmri2bids and BIDS
validator: https://github.com/Squishymedia/bids-validator) and we hope that
mayor pipeline engines (LONI, AA, C-PAC, REST, PSOM, Nipype etc.) will
adopt BIDS.

We have also prepared many example dataset showcasing the new standard:
https://55b6842373553c2422bd99ac533b084533ea6cb7.googledrive.com/host/0B2JWN60ZLkgkQm9ZazZCWS1VNVE/
(they are all task based fMRI, but I will soon add some resting state as
well).

We would love to hear your feedback. Would you recommend this for your lab?
Does it cover the types of studies you do? Either reply to this email or
comment on the current working draft:
https://docs.google.com/document/d/1HFUkAEE-pB-angVcYe6pf_-fVf4sCpOHKesUvfb8Grc/edit?usp=sharing

Best,
Chris Gorgolewski and the INCF Data Sharing Task Force
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.nitrc.org/pipermail/mrtrix-discussion/attachments/20150803/646c0b9a/attachment.html>