ENH: Add csd exercise #117
Conversation
|
Deploy preview for carpentries-dmri ready! Built with commit 615e43f |
|
Deploy preview for carpentries-dmri ready! Built with commit c68dcfa |
There was a problem hiding this comment.
Have not tried c&p-ing the code into a script/notebook to try.
IMHO: the exercise statement is misleading:
In this episode, the ODF of a single fibre population (i.e. no crossing fibres) was visualized.
I'd say that it is inaccurate; the reconstructions, if visualized with enough detail, would show crossing fibers.
If this was meant to refer to the FRF, then we run the risk of generating another confusion by offering learners an exercise that creates fODFs of crossing fibers, which some may think that could be used as a sort of advanced FRF.
So I'd restate the exercise statement.
the file is named odf_multiple_angles.png in the CSD figure directory.
No figure has been added with the commit. Since the jupyter notebook part is still missing, it can done when adding those.
Marking the PR as a draft might be useful until it is complete.
Also, I wouldn't mark #30 as resolved; in #30 it was suggested (unless the wording used was inaccurate or confusing) to perform the actual deconvolution given the peaks for crossing fibers.
Another exercise mentioned would be to actually perform the deconvolution, but I think we need to dig a little deeper into this dataset before that. I think as we expand the episode to include more details spherical harmonics and figure out some of the quirks we are seeing with the dataset, we can add an exercise to deconvolve with different harmonic orders.
Reading this maybe #30 was unclear: it implied using synthetic data (a voxel) (or voxels from the data containing crossings at different angles might do the job) to simulate a signal containing crossings at different angles. I'll add that to the issue.
But I agree, either solution requires more time to investigate and develop in case of the synthetic signal.
Given the time limitations, we may want to go with what this PR proposes.
Fair point, I've restated the sentence to note in the episode we visualized the response function's ODF.
You're right - committed, but forgot to push. Figure should be present now.
Updated the initial PR so it mentions #30 without linking to the issue. |
|
Similar to #120 - going to change this to draft now with plans to sit down one of the nights next week to just solely update this episode with figures, sync content between markdown and notebook(s), run a linter through the notebook and address the changes.. |
Co-authored-by: Michael Joseph <[email protected]>
|
Notebooks and markdown should be synced. Added a solutions notebook for the exercise, and removed the |
ResolvesRelates to #30 - adds an exercise to simulate crossing angles of 90, 60, 45, 30, and 20 degrees. I did have run into some issues simulating withdefault_sphereand ended up usingrepulsion724in the solutions. If someone can double check that, that would be great. I've also left the exercise out of the jupyter notebooks for now. Also have a figure saved showing the solutions at each crossing angle, however can't seem to get it to load in the solutions. If anyone knows how, the file is namedodf_multiple_angles.pngin the CSD figure directory.We can add more exercises later on and also increase the difficulty, but I think this is a good place to start to see where CSD may start to fail.
Another exercise mentioned would be to actually perform the deconvolution, but I think we need to dig a little deeper into this dataset before that. I think as we expand the episode to include more details spherical harmonics and figure out some of the quirks we are seeing with the dataset, we can add an exercise to deconvolve with different harmonic orders.