GSoC 2021 Project Idea 19.2: DIPY project - Population-specific Tractography Bundle Atlas Creation

Description: Implement a method to create a population-specific Tractography Bundle Atlas. Given an input of several subjects’ segmented white matter tracts, create one standard atlas of bundles for the population. The method will utilize the streamline-based registration framework available in DIPY.


  • Understand Diffusion Tensor Imaging and Tractography data
  • Implement Bundle Atlas creation method
  • Write DIPY workflow of the method
  • Test it on different data sets

Difficulty: Hard

Skills required: Python, Registration.

Skills preferred: Experience with Diffusion Tensor Imaging

Mentors: Bramsh Chandio, Shreyas Fadnavis, and Jong Sung Park

Dear mentors, this is David Romero. I am PhD student in biomedical engineering very much interested on the project.
After going through the Direct Bundle Registration tutorial and [Garifallidis,2015] I have some questions:

  1. For clarification, I read how in [Garifallidis,2015] a probabilistic atlas of the optic radiation is built. The final goal of this project would be to do something similar but with other tracts?
  2. Should the atlas building rely only on segmented tracts or would it be possible to use other inputs (e.g. T1w images) to aid the process?
  3. Should the implementation be based entirely on the SLR? As far as I understand SLR uses an affine registration and it might be interesting to introduce non-linear registration. As a possible idea, some authors in my main research field (retinal imaging) use linear registration to create an initial atlas and then use non-linear registration to refine that atlas on an iterative process (I hope it makes sense, I am not so familiar with bundle registration).

Thanks a lot! @bramsh @ShreyasFadnavis @pjsjongsung

1 Like

Hi David,

You are on a right track. I agree with you to include non-linear registration after affine registration. Although, the non-linear registration of streamlines is tricky.

In your proposal, you can add these steps that you think would improve the bundle atlas creation process.

Thanks :+1:, I will think about the overall approach and prepare a draft proposal.

1 Like

Feel free to reach out and share your draft with me at

1 Like

Hi all :love_you_gesture:!

For anyone interested in the project, feel free to visit my Blog, where I will post weekly updates on the progress.

Tagging @malin , @arnab1896 so you can keep track of it.

awesome, thanks! /Malin