[Mrtrix-discussion] query regarding the coordinate system

Sun Apr 29 06:45:33 PDT 2012

Hi Donald,

many thanks for your quick reply.
I understand the importance of using directly the DICOM images in order to avoid/limit problems with the gradient table. I will do that in the future.

Going back to your questions:
1. I am suppliyng manually the gradient table.
2. I have created an enconding file in accordance with mrtrix requirements and passed it to both 'dwi2tensor' and 'streamtrack' commands.

Everything works fine with the original gradient table, that is the corticospinal tract 'dissected' by mrtrix looks biologically plausible. If I check the orientation plots in mrview, as suggested by Luis in the previous email, they look correctly oriented.  If I invert both X an Z axes the output of tractography shows a corticospinal tract which does not look plausible. The orientation plots look also incorrectly orientated.

Based on these empirical results, evidence seems to suggest that the original gradient table should be correct in the particular case of this data set.

Thank you for your precious advice,

Francesco

Il giorno 26/apr/2012, alle ore 12:04, Donald Tournier ha scritto:

Hi Francesco,

I'll just clarify a couple of things before addressing your actual question. You're right that the two coordinate systems differ, but as long as the conventions are clear, there is no problem converting between the two. If you feed DICOM data directly into MRtrix, it will extract the gradient table information from the DICOM headers and convert it into its own coordinate system. So you shouldn't need to worry about it as long as you always use the DICOM data directly, or convert from DICOM into MRtrix's own .mif format (since that will store the gradient table information).

Converting to NIfTI first (as many people seem to do) will introduce problems since the gradient information is not stored in the image header. The issue then is that the user has to supply the gradient table separately, and it's difficult to guarantee that this has been converted into the appropriate coordinate system correctly. For this reason, I strongly recommend to always work directly from the DICOM data if these are available, and to avoid using NIfTI unless the data need to be exported to other packages.

Coming back to your original question, the answer is that MRtrix is fully aware of the DICOM coordinate system convention, and will perform any conversions necessary to ensure consistency with its own coordinate system.

That said, it seems from your original question that you might already be  supplying the gradient table manually. Is this the case? Have you created an encoding file, and passed it to both the dwi2tensor command, and to the streamtrack command (if tracking using the DT_STREAM algorithm)? If so, then I agree this is surprising. However, I would also point out that your description of the coordinate system doesn't match the figure that you supplied - the Z axis increases from superior to inferior, according to the figure. To convert properly, you would need to invert the Z axis as well as the X axis.

The other thing to bear in mind with GE data is that as far as I know, the gradient orientations are actually supplied with respect to the phase/read/slice axes, rather than real/scanner coordinates (which is what the other two major manufacturers do). So it may very well be the case that the original gradient table was indeed OK for this particular data set. However, you need to be extremely careful  in this case since MRtrix expects the gradient orientations to be provided with respect to real/scanner coordinates. This is fine for pure axial scans, since the two coordinate systems will then coincide, but for general oblique imaging planes, it will give the wrong orientations.

So for all the reasons above, I urge you to use MRtrix to convert the DICOM data directly, and to convert the DWI only to .mif format. I've spent a long time making sure that the gradient information could be extracted from the DICOM headers to limit these kinds of problems. Unless you find MRtrix is unable to find the DW directions (i.e. mrinfo <your data> doesn't show a 'DW scheme' entry), then this is by far the safest thing to do.

Hope that helps,

Donald.

On 25 April 2012 10:53, Carletti, Francesco <francesco.carletti at kcl.ac.uk<mailto:francesco.carletti at kcl.ac.uk>> wrote:
Hi all,

I have a query regarding the coordinate frame used by mrtrix.

I have a DTI dataset acquired by a 1.5 Tesla GE SIGNA NVi scanner.
According to the manufacturer the scanner uses a right-handed coordinate system, that is:

dimension description
0 (x) increasing from right to left
1 (y) increasing from posterior to anterior
2 (z) increasing from inferior to superior

I have uploaded the an image showing the details of the coordinate system.
Tractography carried out with mrtrix and the original gradient table (provided by GE) looks correct.

>From the description of the coordinate system used by mrtrix (http://www.brain.org.au/software/mrtrix/general/overview.html#axes) I understand that mrtrix uses a different convention and therefore I think I should swap the X axis.
However, if I do swap X, the results of tractography are not clearly not plausible.

I was wondering if you can shed a light on this.
Many thanks again for your help.

Francesco

<Coordinate_system.png>

Dr Francesco Carletti - Neuroradiologist
Wellcome Research Training Fellow

King's College London
King's Health Partners
Department of Psychosis Studies
Department of Neuroimaging, Center for Neuroimaging Sciences
Institute of Psychiatry
Box PO67
London SE5 8AF, UK

tel:       +44 (0)20 7848 0049
fax:      +44 (0)20 7848 0976
email:  francesco.carletti at kcl.ac.uk<mailto:francesco.carletti at kcl.ac.uk>

_______________________________________________
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

--
Jacques-Donald Tournier (PhD)
Brain Research Institute, Melbourne, Australia
Tel: +61 (0)3 9035 7033

Dr Francesco Carletti
Wellcome Research Training Fellow

Francesco.Carletti at kcl.ac.uk<mailto:Francesco.Carletti at kcl.ac.uk>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.nitrc.org/pipermail/mrtrix-discussion/attachments/20120429/f6ae8b50/attachment.html