[Neuroimaging] JSON-LD and DICOM?

[Matthew Brett]

Hi,

On Fri, Jun 30, 2017 at 5:35 PM, Jasper van den Bosch <japsai at gmail.com> wrote:
> Hi all,
> I have to agree with Andrey on the yet another format argument. Also, tools
> will have to convert it to other formats anyway, so if you do end up storing
> something in the header, as long as you document it, just like you would
> other nibabel properties, I'd go for the simplest solution.

There is going to be a JSON header extension quite soon, so it's not
really another format, but another way of storing metadata in NIfTI.
Then the question is - should we also store DICOM metadata there?

> I did a little survey among niprov users and it turns out they usually don't
> end up using the header to store the metadata, because the "central" storage
> can more easily keep track of the relationships between files.
> And if the file is overwritten, or moved, the metadata remains.

I think the DICOM / NIfTI thing is partly cultural.  If you're a C++ /
cluster sort of person, you tend to prefer DICOM, because there are
C(++) libraries to read the DICOMs, and setting up the libraries /
compiling isn't much of an issue, because everything you need can
easily be set up on the cluster.  If you're a Python / laptop sort of
person, then you tend to prefer NIfTI because it's such a simple
format that it's easy to write your own code to manipulate it, and the
code remains fairly readable, and no extra setup is needed other than
a `pip install pydicom`.

The JSON header extension is also an extremely obvious idea, that we
need to store just a little bit more metadata than we have now.  So -
why not store a lot more metadata than we have now?

Cheers,

Matthew