by

Where communities thrive


  • Join over 1.5M+ people
  • Join over 100K+ communities
  • Free without limits
  • Create your own community
People
Repo info
Activity
  • 14:19
    karanphil commented #2221
  • 14:17

    skoudoro on master

    DOC : move Denoising on its own… Merge pull request #2220 from a… (compare)

  • 14:17
    skoudoro closed #2220
  • 14:16
    skoudoro commented #2220
  • 14:12
    skoudoro milestoned #2221
  • 14:11
    skoudoro labeled #2221
  • 14:11
    skoudoro labeled #2221
  • 14:11
    skoudoro milestoned #2222
  • 14:11
    skoudoro labeled #2222
  • 14:08
    skoudoro commented #2222
  • Sep 19 22:55
    arokem commented #2222
  • Sep 19 22:44
    Redouan20102 opened #2222
  • Sep 18 23:46
    codecov[bot] commented #2206
  • Sep 18 23:42
    pep8speaks commented #2206
  • Sep 18 23:42
    CHrlS98 synchronize #2206
  • Sep 18 23:34
    pep8speaks commented #2206
  • Sep 18 23:34
    codecov[bot] commented #2206
  • Sep 18 23:34
    CHrlS98 synchronize #2206
  • Sep 18 16:08
    codecov[bot] commented #2206
  • Sep 18 16:08
    pep8speaks commented #2206
Gabriel Girard
@gabknight
@imanusita yes, you can control this with max_angle and the step_size for both deterministic and probabilistic tracking. try using the deterministicMaxDG with max_angle=20 and step_size=0.2.
imanusita
@imanusita
so can u plz tell me what is the advantage of using the probabilistic algorithm over using the deterministic
Gabriel Girard
@gabknight
@imanusita there are many considerations; it is application dependent. This is discuss in the literature, for instance see https://onlinelibrary.wiley.com/doi/full/10.1002/nbm.3785 and reference therein.
imanusita
@imanusita
owkey thnk u
can u plz tell me which algorithm u used for probabilistic tractography i mean (like fact tend ...) and for deterministic sorry for my silly questions but i m preparing to my defense and i m waiting such questions . thnk u again
Gabriel Girard
@gabknight
imanusita
@imanusita
ya ya i read it i just wanna know what is the algo used by default in ur platform
Gabriel Girard
@gabknight
there is no default, we use both probabilistic (e.g. probabilisticdirectiongetter) and deterministic (e.g. deterministicmaximumdirectiongetter) algorithms.
imanusita
@imanusita
owkey thnk u ^^
Ewa
@EwaRowe91_gitlab
HI, I need help in understanding what algorithm is under set_number_of_points from dipy.tracking.streamline. Thank you in advance for your answers.
Ariel Rokem
@arokem
Paolo Avesani
@Paolopost
A question on dipy_track_local workflow. Should I expect that the time of computation is approximately 8x using "seed_density" parameter 2 rather than 1 (i.e. seeding 8 points per voxel instead of 1)?
Eleftherios Garyfallidis
@Garyfallidis
@Paolopost correct.
Paolo Avesani
@Paolopost
A suggestion to improve the dipy_track_local workflow. Currently there is no opportunity to set the step-size parameter of tracking. It is fixed to 0.5mm. May be it would be worthwhile at least to set this value proportionally to the voxel size, e.g. half of it.
Serge Koudoro
@skoudoro
Thank you @Paolopost for the suggestion. This has been already added on master so it will be there for next release !
Yijun Liu
@snapfinger
Hi guys, does anyone have unprocessed diffusion weighted images of kids and acquired with 64 diffusion directions per shell (uniform distribution)? Didn't find suitable data in HCP/NITRC/OpenNeuro, if you have we can probably collaborate. And would appreciate if anyone can point to the link if you know there's such public dataset.
Michael Paquette
@mpaquette
@wayalan Might not be relevant anymore, as your question dates from April 11th, but using the G/DELTA/delta value you provided, I am estimating a b-value of 6703.35 s/mm^2, close enough to the 6650 reported by your scanner. You might be using wrong unit conversion or the wrong value of gyromagnetic ratio (should be ~(42.515e6 x 2pi) rad T^-1 s^-1)
Eleftherios Garyfallidis
@Garyfallidis
An alert here to say that we are making a lot of refactoring to accommodate the upcoming DIPY 1.0 release. If you are using dev version, please use caution.
Serge Koudoro
@skoudoro
FYI, We plan to cut the release this Wednesday. (31rst)
Daniel Haehn
@haehn
Hello Dipy Devs! I am getting involved in a diffusion project and I have the following question: How do you store streamlines and attached scalars and properties? We are running into limitations of the TrackVis file format and the scientists here so far used VTP - everything less than ideal. How do you do it? I saw dipy also exports TRK files.
Francois Rheault
@frheault

@haehn I am working on example for the new stateful_tractogram class made to facilitate loading/saving various file format (including trk)
The easiest way in the meantime to get an example is the test in https://github.com/nipy/dipy/blob/master/dipy/io/tests/test_stateful_tractogram.py and look for ''random_streamline_color'' or ''random_point_gray'' or ''random_point_color''.

That would cover data_per_point. For data_per_streamline it is easier, the new class accept a dictionnary. The key is the name is name of your metadata and the value is an array or a list and then len() must be the same as the number of streamlines.

Also this new class require a reference so make sure you have the nifti file they were generated with and know in which space they are when creating the object.
(If you have no idea what I meant, tell me)

Daniel Haehn
@haehn
Thanks. But which file format do you ultimately choose? TrackVis is too limited at this point in regards to scalars and properties.
VTP files are too large.
is DPY preferred?
Eleftherios Garyfallidis
@Garyfallidis
Hi @haehn, great to hear from you. Eventually DPY should be the preferred option. However, we are not promoting it much for now as we want to add some crucial new options. Can you give us an example of your user case?
Daniel Haehn
@haehn
Thank you, @Garyfallidis ! The use case is standard fiber bundles with flexible numbers of scalars (per fiber point) and flexible number of properties (per fiber). TRK files limit it to 10 each.
Is that supported yet with DPY?
Eleftherios Garyfallidis
@Garyfallidis
@haehn we can easily have support for any number of properties in DPY. However, not there yet. I will try to prioritize it so that others like you can be facilitated. Alternatively, you can create a very fast format with NPZ using the Streamlines API that we developed in Nibabel and use DIPY. Basically a Streamlines object can give you the arrays of the points and the metrics (of any number - look at the data_per_point attributes/parameters) as numpy arrays.
Eleftherios Garyfallidis
@Garyfallidis
Look at the parameters that a Tractogram object can have data_per_point and per_streamline. You can then save these data as independent arrays and reconstruct the Tractogram on load.
Erin Teich
@erteich
Hi all! I'm a fairly new user of dipy, and I've been doing some playing around with the HARDI labeled dataset accessible via dipy.data.read_stanford_labels(). I saw that this is relabeled, such that white matter voxels have the label 1 or 2, regardless of what labels freesurfer originally gave them. I was wondering if it is possible to gain access to the original freesurfer labels for these voxels, however they were reduced from the original aparc+aseg file? Then I could delineate different white matter regions according to that parcellation. Thanks in advance for the help!
Eleftherios Garyfallidis
@Garyfallidis
And don't forget to save the streamlines too. From them Tractogram.streamlines you will need the data, offsets and lengths arrays to be saved in the NPZ (numpy compressed) format. It's an idea. Let me know if you try it. Right now we are busy with the upcoming release but building a tutorial with these issues and updating the DPY format should be an easy and fun project.
Eleftherios Garyfallidis
@Garyfallidis
@haehn issue added here nipy/dipy#1936 and targeted for release 1.1 aimed in October.
Daniel Haehn
@haehn
Thank you so much @Garyfallidis ! I will follow the DPY issue and also try the NPZ.. then I will compare file size against the TRK and VTP formats.
Casa Mofoekeng 'Moso
@CasaMofoekengMo_twitter
Hello, I am trying to implement Python code when given the names and GPS positions of 750 people (latitude, longitude and elevation) to find the names of the 10 closest neighbors of a randomly selected individual.
please help.
Eleftherios Garyfallidis
@Garyfallidis
This is not the right channel for this question @CasaMofoekengMo_twitter but hey you can google KNN and python I am sure they will be many solutions available. Best of luck.
Eleftherios Garyfallidis
@Garyfallidis
A historic moment. DIPY 1.0 is out!
Serge Koudoro
@skoudoro
:thumbsup: :punch: :fire: :rocket:
Bramsh Q Chandio
@BramshQamar

A historic moment. DIPY 1.0 is out!

:thumbsup: :finnadie:

Eleftherios Garyfallidis
@Garyfallidis
:fire: :fire: :fire: :fire: :fire: :fire: :fire:
:roller_coaster: :roller_coaster: :roller_coaster:
Bramsh Q Chandio
@BramshQamar
:fireworks:
Eleftherios Garyfallidis
@Garyfallidis
:rocket: :rocket: :rocket: :rocket: :rocket: :rocket:
Shreyas Fadnavis
@ShreyasFadnavis
👌🏻👌🏻👌🏻👌🏻👌🏻🔥
Gabriel Girard
@gabknight
:thumbsup: :thumbsup:
Rafael Neto Henriques
@RafaelNH
Congrats to all ! Well done! Amazing !
Romulus
@romainviard
Thx u !! in hurry to try it
Paolo Avesani
@Paolopost