[Mrtrix-discussion] Specifying gradient directions on a GE Signa HDx scanner

[Winston, Gavin]

After discussion with Donald, it appears that the sequence in use on our GE scanner (a research sequence) does not write the gradient directions to the DICOM file so there is no way to extract them from the DICOM.  I have always previously used a manual gradient table for Camino / FSL.  Since Camino uses its own voxel order format, the rotation does not matter as it assumes the gradients are relative to the acquisition plane.  Similarly from my experiments, FSL seems to apply the same logic as it does not make any difference what is in the qform/sform header.

However MRtrix does use the transforms in the header so when you tilt the acquisition, the fitting is incorrect unless you change the transform to an identity as I suggested or rotate the gradient vectors.  Using the identity transform is a quick fix but is not ideal as some of the other processing I do with the same relies on these being present.  So does anybody have a ready-made script which takes a set of gradient directions and rotates them according to the qform/sform?  I seem to remember seeing something like this recently but cannot find in on the MRtrix list so it must have been elsewhere.

Many thanks

Gavin


From: Donald Tournier [mailto:d.tournier at brain.org.au]
Sent: 16 October 2012 00:18
To: Winston, Gavin
Cc: mrtrix-discussion at www.nitrc.org
Subject: Re: [Mrtrix-discussion] Specifying gradient directions on a GE Signa HDx scanner

Hi Gavin,

Yes, the fun never ends with DICOM...

MRtrix was actually coded up based on the same wiki information, but I only implemented support for the latest version, as per the last few entries in the wiki:

In GE DWI images (software version 12.0)





(0008,1090) LO [SIGNA EXCITE]

(0018,1020) LO [12\LX\MR Software release:12.0_M4_0520.a]

diffusion directions are stored under the following tags:





(0019,10bb) DS [0.430617]

(0019,10bc) DS [-0.804161]

(0019,10bd) DS [-0.420008]

This has worked fine for all the GE data I've had access to (admittedly not a lot), and it does take care of re-orienting the directions into scanner coordinates. I guess it would be good to include support for these other data, especially if this is actually the GE 'standard'. I'd be happy to have a go if you can send me some data to try it out. It sounds like a relatively trivial change, I just need to parse the (0021,105A) tag... Actually, just a dump of the output of "read_dicom -a <DICOM file>" should be sufficient for me to figure out how it's stored. If you can send me that, I'll add support for it.

Otherwise, the directions you get from the headers sound like they would be similar to the FSL bvecs/bvals format. There's been a bit of discussion about how to convert between FSL & MRtrix formats recently, with a few people contributing their scripts. A quick google search for MRtrix & bvecs should give you the info you need.

Finally, I see no reason why you couldn't use the trick of resetting the transform matrix to identity, unless you actually need to preserve the original orientation.

Hope this helps.
Cheers,

Donald.


On 15 October 2012 23:08, Gavin Winston <g.winston at ucl.ac.uk<mailto:g.winston at ucl.ac.uk>> wrote:
Dear all,

I would like to use MRtrix on DTI data from a GE Signa HDx 3.0T scanner (software release 14.0).  My query relates to how to specify the gradient directions in the correct reference frame for MRtrix.

The gradient table file on the scanner specifies the three components of the diffusion gradients relative to the acquisition (i.e. if you tilt the acquisition plane by 10 degrees, the applied gradients also tilt by 10 degrees).  I think this behaviour differs from Siemens scanners.  If I use the gradient table directly in MRtrix and the acquisition is tilted the fibre orientations are also incorrectly tilted.  There seem to be several ways to approach this so which do people advise?

Firstly, extracting the gradient directions from the DICOM header.  I understand from previous posts that GE scanners are a bit funny about how they do this.  Using read_dicom on our data seems to suggest that it is stored in the fields 0019,10E0 (DTI diffusion directions), 0021,105A (diffusion direction), 0043,1039 (b value) as per http://wiki.na-mic.org/Wiki/index.php/NAMIC_Wiki:DTI:DICOM_for_DWI_and_DTI#Private_vendor:_GE.  However mrinfo does not seem to support this as it just produces a gradient table which is entirely zeros.

Secondly, I could reorientate the entries of the gradient table according to the transforms from the DICOM file (or the qform/sform once converted to Nifti).  Are there are ready-made tools to do this?

Finally, it is possibly to implement a quick fix by modifying the qform/sform headers.  If you convert DICOM -> NIFTI, using fslorient as below then convert NIFTI -> MIF with mrconvert things work fine.  For example on a 1.875x1.875x2.4mm dataset:

fslorient -setqform -1.875 0 0 0 0 1.875 0 0 0 0 2.4 0 0 0 0 1 <file>
fslorient -setsform -1.875 0 0 0 0 1.875 0 0 0 0 2.4 0 0 0 0 1 <file>

Normally our acquisitions are virtually aligned with the scanner axes so the error is small, but I acquired a dataset this morning where the acquisition was tilted by 45 degrees to check the above worked.

Many thanks,

Gavin
_______________________________________________
Mrtrix-discussion mailing list
Mrtrix-discussion at www.nitrc.org<mailto:Mrtrix-discussion at www.nitrc.org>
http://www.nitrc.org/mailman/listinfo/mrtrix-discussion



--
Dr Jacques-Donald Tournier
Research Fellow

The Florey Institute of Neuroscience and Mental Health
Melbourne Brain Centre - Austin Campus
245 Burgundy Street
Heidelberg  Vic  3084
Ph:  +61 3 9035 7033
Fax:  +61 3 9035 7307
www.florey.edu.au<http://www.florey.edu.au>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.nitrc.org/pipermail/mrtrix-discussion/attachments/20121017/49b44304/attachment-0001.html