search

bids-standard / bids-specification

Brain Imaging Data Structure (BIDS) Specification
https://bids-specification.readthedocs.io/
Creative Commons Attribution 4.0 International
272 stars 156 forks source link

Support for susceptibility imaging including QSM, SWI and SWI-MIP #1566

Open astewartau opened 1 year ago

astewartau commented 1 year ago

Your idea

Support for Quantitative Susceptibility Mapping (QSM) currently exists as it is considered a quantitative, derivative and anatomical MR image and uses the _Chimap suffix. It also has a BIDS example. I believe this was introduced by the qMRI extension integrated in 2022 based on Karakuzu et al.'s Scientific Data publication.

However, similar support for Susceptibility Weighted Imaging (SWI) and associated Minimum Intensity Projections (SWI-MIP) is to be integrated. The inactive BEP004 proposal suggests a swi/ datatype and proposes storing the input _GRE images there with the resultant files using proposed _swi and _minIP suffixes.

Currently, _MEGRE and _T2starw suffixes already exist in the anat/ datatype, so the remaining changes to fully support SWI may be smaller and consist only of adding the _swi and _minIP suffixes.

Myself, @stebo85, and @neurolabusc propose to finalise support for susceptibility imaging by making some minor pull requests to do this. We also have open examples on OSF of DICOM input data and its mapping to BIDS, which are suitable for QSM and SWI reconstruction that we propose to adapt into similar BIDS examples.

We are opening this issue to begin discussions with the wider susceptibility imaging and software developer community so we can share ideas before changes are made and integrated.

tsalo commented 1 year ago

I believe you will also need a coil or channel entity. There's an open PR for that (#425), but it stalled a while back (mostly my fault).

stebo85 commented 1 year ago

Dear @tsalo,

Although storing single channel data was something we had to do in the past to get proper phase combination offline, this is now solved quite well by proper channel combinations on the systems that deliver phase without singularities.

Would there be a tool that needs single channel data as an input which would really profit from this? My feeling is that adding this would complicate things without providing much benefit?

Even if single channel information would be seen as useful in BIDS, I would argue that it shouldn't be part of the SWI/QSM proposal, because it would affect also other modalities and would be something more general?

Thank you Steffen

tsalo commented 1 year ago

@stebo85, you definitely would know more about SWI than I do. The coil entity was part of the original SWI BEP, and I only broke out that element into a separate PR because I had some single-channel data on hand. If online coil combination methods are suitable at this point, then you can definitely drop the coil entity.