I can compute the voxel-wise coefficients of the corresponding spherical
harmonics fit of the diffusion-weighted signal using equation [image:
Inline image 2] as given in the research paper of analytical QBI paper.
Afterwards I am confused how to sample these coefficients to 100 directions
equally distributed over the hemisphere? I can estimate SH basis for for
100 directions but how to link this information to the computed
coefficients?
Tabinda
On Thu, Jan 25, 2018 at 5:29 PM, Neher, Peter <p.ne...@dkfz-heidelberg.de>
wrote:
> If you go to the “Q-Balls” view in MITK Diffusion, you can also select
> “Raw Signal Only” as reconstruction and you will simply get an
> interpolation of the input signal, albeit with a different number of
> sampling directions. This is maybe good for visualizing what the result
> looks like.
>
>
>
> *Von:* Neher, Peter [mailto:p.ne...@dkfz-heidelberg.de]
> *Gesendet:* Donnerstag, 25. Januar 2018 07:26
> *An:* Tabinda Sarwar <tsar...@student.unimelb.edu.au>;
> mitk-users@lists.sourceforge.net
>
> *Betreff:* Re: [mitk-users] Resampling the signal for tractography using
> machine learning
>
>
>
> Hi Tabinda,
>
>
>
> this is indeed a bit confusing. We do not calculate actual Q-balls here.
> The analytical qball filter is also capable of only fitting the raw signal
> with spherical harmonics. This is done by setting “filter->
> SetNormalizationMethod(InterpolationFilterType::QBAR_RAW_SIGNAL);”
>
>
>
> By the way, please answer to the list so that other users can also read
> out conversation.
>
>
>
> Peter
>
>
>
> *Von:* Tabinda Sarwar [mailto:tsar...@student.unimelb.edu.au
> <tsar...@student.unimelb.edu.au>]
> *Gesendet:* Donnerstag, 25. Januar 2018 00:38
> *An:* Neher, Peter <p.ne...@dkfz-heidelberg.de>
> *Betreff:* Re: [mitk-users] Resampling the signal for tractography using
> machine learning
>
>
>
> Hi Peter,
>
>
>
> The high level details which I can understand is that first from raw input
> dMRI ODF is computed using analytical Q-ball method which undergoes
> sampling. If I am right, doesn't it makes the input features dependent on
> the ODF computation technique?
>
>
>
> Tabinda
>
>
>
> On Wed, Jan 24, 2018 at 5:45 PM, Neher, Peter <p.ne...@dkfz-heidelberg.de>
> wrote:
>
> Dear Tabinda,
>
> we interpolate the original signal using spherical harmonics and then
> sample 100 directions form this representations:
> https://phabricator.mitk.org/source/mitk/browse/master/
> Modules/DiffusionImaging/FiberTracking/Algorithms/TrackingHandlers/
> mitkTrackingHandlerRandomForest.cpp;2642ee4902f4b1e8f5c443f22b30aa
> 896795d5da$105
>
> The actual fitting of the spherical harmonics and the sampling is done in
> this class: https://phabricator.mitk.org/source/mitk/browse/master/
> Modules/DiffusionImaging/DiffusionCore/include/Algorithms/Reconstruction/
> itkAnalyticalDiffusionQballReconstructionImageFilter.h
>
> The number of directions is defined by the template parameter
> NumberOfSignalFeatures of the TrackingHandlerRandomForest class.
>
> Cheers,
> Peter
>
> ---------------------------------------------------------------------
>
> Dr. Peter F. Neher
> Division of Medical Image Computing (MIC)
> Scientist
>
> German Cancer Research Center (DKFZ)
> Im Neuenheimer Feld 280
> <https://maps.google.com/?q=Im+Neuenheimer+Feld+280+%0D+69120+Heidelberg+%0D+Germany&entry=gmail&source=g>
> 69120 Heidelberg
> Germany
> Phone: +49 6221 42-2330
>
> p.ne...@dkfz.de
> www.dkfz.de/en/mic/
>
> ________________________________________
> Von: tabinda <tsar...@student.unimelb.edu.au>
> Gesendet: Mittwoch, 24. Januar 2018 06:16
> An: mitk-users@lists.sourceforge.net
> Betreff: [mitk-users] Resampling the signal for tractography using machine
> learning
>
>
>
> Hi!
>
> I am reading the paper "Neher, P. F., Côté, M.-A., Houde, J.-C.,
> Descoteaux,
> M. & Maier-Hein, K. H. Fiber tractography using machine learning.
> Neuroimage
> 158, 417–429 (2017)" and am trying to understand the methodology using the
> MITK source code. But I am confused how the input raw signal is resampled
> to
> 100 directions? I couldn't locate the source code for this particular step.
> I need to know how mathematically this is done?
> For example if original data is (140x140x90) is acquired using 30 gradients
> (which makes the dMRI 140x140x90x30), resampling causes the dMRI to be of
> size 140x140x90x100? Guidance on the resampling will help me to understand
> the algorithm. Any help will be highly appreciated.
>
> Thanks and Regards,
> Tabinda
>
>
>
> --
> Sent from: http://mitk-users.1123740.n5.nabble.com/
>
> ------------------------------------------------------------
> ------------------
> Check out the vibrant tech community on one of the world's most
> engaging tech sites, Slashdot.org! http://sdm.link/slashdot
> _______________________________________________
> mitk-users mailing list
> mitk-users@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/mitk-users
>
>
>
------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot
_______________________________________________
mitk-users mailing list
mitk-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/mitk-users
