[Mrtrix-discussion] First voxel of track is not a seed voxel

[Michael Schirner]

Hi Donald,

thank you for this information and for providing a fix so quickly - this
helps a lot!
Indeed we need to make sure that there are only tracks emerging from the
exact seeding voxels. Since the problem occurs only in some 15 % of all
tracks we sorted them out using a small script, but your update is of
course the better solution and we'll use that from now on.

atb,
Michael


> Hi Michael,
>
> OK, I figured out what the issue is. The -nomaskinterp only applies to
> include, exclude & mask ROIs, but not to the seed ROI (it's a different
> section of code that handles this bit). The current seeding strategy is to
> find a point at random where the intensity in the mask is higher than 0.5.
> When using tri-linear interpolation, if you have a zero voxel with a few
> 'one' voxels around one of its corner, some of the zero voxel's volume
> will
> end up with intensity higher than 0.5 - just enough to allow a few seed
> points to appear within a voxel that wasn't actually part of the mask.
>
> It's nothing to worry about in most cases, since the spurious seeds will
> be
> located no more than a fraction of a voxel outside the ROI. However, if
> you
> really need to ensure that the seed points are strictly located within the
> ROI's voxels, then obviously it's a problem. If you need to get rid of
> that
> problem, I've just committed a fix for it, which makes sure the
> -nomaskinterp option also applies to the seed ROI. You can get it direct
> from the SVN repo on Google Code:
> https://code.google.com/p/mrtrix/source/checkout
>
> Hope this helps!
> Cheers,
>
> Donald.
>
>
>
>
> On 23 January 2014 17:52, Michael Schirner <m.schirner at fu-berlin.de>
> wrote:
>
>> Hi Donald,
>>
>> thanks for coming back to me. This post was a follow up to the post I
>> posted yesterday, I didn't manage to add it to the thread.
>>
>> Below is the text of yesterdays post. I'm happy to provide you with the
>> data if that helps.
>>
>> atb,
>> Michael
>>
>> ***snip***
>> Hi Donald et al,
>>
>> we encounter some strange behaviour: streamtrack produces very nice
>> tracks, but it is often the case, that the first voxel of a track is not
>> one of the specified seed voxel but right next to one (despite using the
>> 'unidirectional' flag).
>>
>> We run streamtrack using this command:
>> streamtrack SD_PROB CSD8.mif -seed seedmask.mif -include targetmask.mif
>> -mask mask.mif -nomaskinterp -unidirectional -num 1000 tracks.tck
>>
>> The dwi data was read in from a DICOM directory, no further processing
>> except as specified in the mrtrix documentation. All masks were
>> generated
>> from freesurfer segmentations, stored as NIFTI and converted to
>> mif-format
>> using mrconvert. After tracking we convert the tck files to vtk files
>> using one of the mask-files as '-voxel' image. We tried it with and
>> without adding +1 to each coordinate dimension, in order to get
>> matrix-cube indices.
>>
>> Strangely, when we use masks that contain only a single seed voxel,
>> everything works just fine, but as soon if we use a larger ROI we
>> encounter this problem.
>>
>> Any help is much appreciated. Thanks!
>>
>> Michael
>> ***
>>
>> > Hi Michael,
>> >
>> > I'm not too sure what you mean, so this is just a guess: unless you
>> > supplied the -unidirectional option to streamtrack, that is the
>> expected
>> > behaviour - by default, tracks will proceed in opposite directions
>> from
>> > the
>> > seed point, so that it ends up somewhere in the middle of the track.
>> >
>> > Hope that answers the question. If I've got the wrong end of the
>> stick,
>> > please provide a bit more context...
>> >
>> > Cheers,
>> > Donald
>> >
>> > --
>> > Dr J-Donald Tournier (PhD)
>> >
>> > Senior Lecturer, Biomedical Engineering
>> > Division of Imaging Sciences & Biomedical Engineering
>> > King's College London
>> >
>> > A: Department of Perinatal Imaging & Health, 1st Floor South Wing, St
>> > Thomas' Hospital, London. SE1 7EH
>> > T: +44 (0)20 7188 7118 ext 53613
>> > W:
>> >
>> http://www.kcl.ac.uk/medicine/research/divisions/imaging/departments/biomedengineering
>> >
>> > On 23 Jan 2014 12:38, "Michael Schirner" <m.schirner at fu-berlin.de>
>> wrote:
>> >
>> >> The problem seems to improve when using high-res masks (1mm
>> isotropic).
>> >> In
>> >> this case only a smaller fraction of seed-voxels is affected. Seems
>> like
>> >> there occurs a rounding error at some point. Nevertheless strange
>> that
>> >> it
>> >> never occurs when using only single-voxel masks.
>> >>
>> >> _______________________________________________
>> >> Mrtrix-discussion mailing list
>> >> Mrtrix-discussion at www.nitrc.org
>> >> http://www.nitrc.org/mailman/listinfo/mrtrix-discussion
>> >>
>> >
>>
>>
>>
>
>
> --
> *Dr J-Donald Tournier (PhD)*
>
> *Senior Lecturer, **Biomedical Engineering*
>
> *Division of Imaging Sciences & Biomedical EngineeringKing's College
> London*
>
>
> *A: Department of Perinatal Imaging & Health, 1st Floor South Wing, St
> Thomas' Hospital, London. SE1 7EH*
> *T: +44 (0)20 7188 7118 ext 53613*
> *W:
> http://www.kcl.ac.uk/medicine/research/divisions/imaging/departments/biomedengineering
> <http://www.kcl.ac.uk/medicine/research/divisions/imaging/departments/biomedengineering>*
>