A few tractometry functions #1695

arokem · 2018-12-17T23:16:44Z

For tractometry, we want to be able to extract the profiles along the length of a bundle

This PR introduces a couple of useful functions for that:

bundle_profile takes a bundle and returns the values along the length of the bundle, resampled to 100 (or some other number of) nodes.
gaussian_weights allows you to weight the streamlines within a bundle, depending on how far they are from the core of the bundle.
Finally, orient_by_streamline will make sure that a bundle is all oriented in the same direction (using the other functions without doing this will result in erroneous results!), based on the trajectory of a "standard" streamline.

Thanks to @jyeatman for guidance in implementing these in the prototype implementation we are using in pyAFQ!

skoudoro

Thank you for this @arokem. I just have a few request:

Can you rebase this PR?
Can you add a short tutorial here?
I think it will be better to avoid list of array promotion and use Streamlines class everywhere. I do not understand why you convert your Streamlines object into a np.array. It should be the opposite.

dipy/tracking/streamline.py

skoudoro · 2019-01-03T17:49:58Z

dipy/tracking/streamline.py

+    Parameters
+    ----------
+    data : 3D volume
+    bundle : StreamLines class instance, list of arrays, or array


due to memory issue, I think we should avoid to promote list of arrays and only use Streamlines class instance.

OK. Changed here and in the docstring above.

dipy/tracking/streamline.py

arokem · 2019-01-04T20:02:12Z

I am trying to create an example, which would follow the bundle extraction example, using these bundles to do tractometry. For that, I will also need parameter maps for this subject.

@Garyfallidis: Which of the HCP subjects was used to create the fetch_target_tractogram_hcp tractogram?

arokem · 2019-01-04T20:03:44Z

Sorry - @BramshQamar : I see that you made that tutorial. Do you know which subject fetch_target_tractogram_hcp belongs to?

BramshQamar · 2019-01-05T04:35:00Z

Sorry - @BramshQamar : I see that you made that tutorial. Do you know which subject fetch_target_tractogram_hcp belongs to?

Hello @arokem, we used subject 100206 from HCP data to create target tractogram for bundle extraction example.

codecov-io · 2019-01-05T21:51:34Z

Codecov Report

❗ No coverage uploaded for pull request base (master@61dbe1d). Click here to learn what that means.
The diff coverage is 95.94%.

@@            Coverage Diff            @@
##             master    #1695   +/-   ##
=========================================
  Coverage          ?   84.29%           
=========================================
  Files             ?      115           
  Lines             ?    13748           
  Branches          ?     2176           
=========================================
  Hits              ?    11589           
  Misses            ?     1653           
  Partials          ?      506

Impacted Files	Coverage Δ
dipy/data/__init__.py	`82.05% <ø> (ø)`
dipy/segment/bundles.py	`91.8% <100%> (ø)`
dipy/data/fetcher.py	`33.59% <100%> (ø)`
dipy/align/streamlinear.py	`87.17% <100%> (ø)`
dipy/tracking/streamline.py	`70.73% <91.17%> (ø)`

pep8speaks · 2019-01-05T21:51:36Z

Hello @arokem, Thank you for updating !

In the file dipy/data/fetcher.py, following are the PEP8 issues :

Line 453:9: E128 continuation line under-indented for visual indent

In the file dipy/tracking/streamline.py, following are the PEP8 issues :

Line 723:16: E222 multiple spaces after operator

In the file dipy/tracking/tests/test_utils.py, following are the PEP8 issues :

Line 338:37: E128 continuation line under-indented for visual indent
Line 339:37: E128 continuation line under-indented for visual indent

In the file doc/examples/afq_tract_profiles.py, following are the PEP8 issues :

Line 43:1: E402 module level import not at top of file
Line 63:1: E402 module level import not at top of file
Line 64:1: E402 module level import not at top of file
Line 71:1: E402 module level import not at top of file
Line 73:1: E402 module level import not at top of file
Line 100:1: E402 module level import not at top of file
Line 111:81: E501 line too long (84 > 80 characters)
Line 117:1: E402 module level import not at top of file
Line 126:1: E402 module level import not at top of file
Line 173:4: W292 no newline at end of file

In the file doc/examples/bundle_extraction.py, following are the PEP8 issues :

Line 132:42: E231 missing whitespace after ','
Line 132:45: E231 missing whitespace after ','
Line 133:47: E231 missing whitespace after ','
Line 133:50: E231 missing whitespace after ','
Line 157:10: E201 whitespace after '('
Line 178:43: E231 missing whitespace after ','
Line 178:46: E231 missing whitespace after ','
Line 179:48: E231 missing whitespace after ','
Line 179:51: E231 missing whitespace after ','
Line 215:4: W292 no newline at end of file

Comment last updated on February 28, 2019 at 03:44 Hours UTC

dipy/tracking/streamline.py

arokem · 2019-01-06T04:15:52Z

I added an example usage in f727bde. I had to introduce some changes in the bundle extraction example, because I am reusing the bundles generated in that example.

skoudoro

Thank you for the tutorial! I need to play a bit more with it but in overall, It looks good to me.

dipy/tracking/streamline.py

doc/examples/bundle_extraction.py

skoudoro · 2019-01-08T01:21:41Z

dipy/tracking/streamline.py

+        # This should come back as a 3D covariance matrix with the spatial
+        # variance covariance of this node across the different streamlines
+        # This is a 3-by-3 array:
+        node_coords = bundle.data[node::n_points]


[node::n_points] Nice! 👍

skoudoro · 2019-01-14T16:55:55Z

After a rebase, this PR is ready to go, it will be good if someone could have a look on this one

arokem · 2019-01-14T17:19:34Z

Now rebased!

…

On Mon, Jan 14, 2019 at 8:55 AM Serge Koudoro ***@***.***> wrote: After a rebase, this PR is ready to go, it will be good if someone could have a look on this one — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#1695 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAHPNmqyyHZzBPIOvrp2Hx_zC_VtgfVhks5vDLabgaJpZM4ZXTgw> .

arokem · 2019-01-22T06:31:11Z

Any more thoughts here?

arokem · 2019-01-26T20:23:59Z

Rebased again, after the #1652 merge.

skoudoro

I will merge this on Wednesday if there is no other comment.

Can you fix the small rebase residue below @arokem?

skoudoro · 2019-02-01T15:27:13Z

doc/examples_index.rst

@@ -167,6 +167,10 @@ Streamline analysis and connectivity
 - :ref:`example_streamline_length`
 - :ref:`example_cluster_confidence`
 - :ref:`example_path_length_map`
+<<<<<<< HEAD


rebase residue

skoudoro · 2019-02-01T15:27:28Z

doc/examples_index.rst

+<<<<<<< HEAD
+=======
+- :ref:`example_bundle_profiles`
+>>>>>>> Adds an example of bundle profile extraction (tractometry)


rebase residue

arokem · 2019-02-01T15:36:23Z

Oh - sorry about that - should be fixed now.

…

On Fri, Feb 1, 2019 at 7:29 AM Serge Koudoro ***@***.***> wrote: ***@***.**** commented on this pull request. I will merge this on Wednesday if there is no other comment. Can you fix the small rebase residue below @arokem <https://github.com/arokem>? ------------------------------ In doc/examples_index.rst <#1695 (comment)>: > @@ -167,6 +167,10 @@ Streamline analysis and connectivity - :ref:`example_streamline_length` - :ref:`example_cluster_confidence` - :ref:`example_path_length_map` +<<<<<<< HEAD rebase residue ------------------------------ In doc/examples_index.rst <#1695 (comment)>: > @@ -167,6 +167,10 @@ Streamline analysis and connectivity - :ref:`example_streamline_length` - :ref:`example_cluster_confidence` - :ref:`example_path_length_map` +<<<<<<< HEAD +======= +- :ref:`example_bundle_profiles` +>>>>>>> Adds an example of bundle profile extraction (tractometry) rebase residue — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#1695 (review)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAHPNutWecKLPDMpOyUHAeeJc6vv3pbgks5vJF1mgaJpZM4ZXTgw> .

arokem · 2019-02-12T21:11:45Z

Does anyone else want to take a look? Any reason not to merge this?

skoudoro · 2019-02-12T21:18:53Z

ping @Garyfallidis

arokem · 2019-02-13T18:27:59Z

Looks like a new release of cvxpy is breaking MAPMRI.

skoudoro · 2019-02-13T20:15:43Z

I just did a quick check on cvxpy changelog and there are very few changes. Mapmri tests seems really sensitive... I will have a deeper look later

skoudoro · 2019-02-14T17:51:06Z

Hi @arokem, I just created this issue https://github.com/cvxgrp/cvxpy/issues/672 for our cvxpy problem. Concerning Travis, I do not know what should be our strategy to make Travis happy until their fix

arokem · 2019-02-14T17:57:27Z

Pin the cvxpy version to the previous release for now?

arokem · 2019-02-14T18:01:15Z

Implemented in cbf9275, I think.

skoudoro · 2019-02-14T18:03:50Z

Great! Thanks! we have to put a reminder somewhere for coming back to normal later

See: https://github.com/cvxgrp/cvxpy/issues/672

This reverts commit cbf9275.

Also, add references, fix typo in documentation.

@Garyfallidis

Also, multiple other changes according to review by @Garyfallidis.

arokem · 2019-02-27T04:21:39Z

Now also rebased.

Garyfallidis · 2019-02-27T22:04:07Z

@arokem ready for final check?

arokem · 2019-02-27T22:06:49Z

I believe so!

Garyfallidis · 2019-02-27T22:08:32Z

dipy/segment/bundles.py

+    return w / np.sum(w, 0)
+
+
+def afq_tract_profile(data, bundle, affine=None, n_points=100,


You are already in bundles.py no need for tract. I would just say afq_profile.

Garyfallidis · 2019-02-27T22:10:10Z

I added a last comment. Please correct and ping me to merge! :)

Garyfallidis · 2019-02-27T22:16:11Z

In general I use bundles to mean tracts, fascicles, muscle fibers etc. The tools are quite generic so I think bundles is good name to use. I don't think we need to specify tracts anymore. However, in the tutorial it is good to have some anatomical naming standards. Which you do 👍

arokem · 2019-02-27T22:45:34Z

Yeah - good call. Renamed!

skoudoro reviewed Jan 3, 2019

View reviewed changes

arokem force-pushed the bundle_profiles branch from 07cfef0 to b938a61 Compare January 6, 2019 03:22

arokem commented Jan 6, 2019

View reviewed changes

dipy/tracking/streamline.py Outdated Show resolved Hide resolved

skoudoro reviewed Jan 7, 2019

View reviewed changes

skoudoro reviewed Jan 8, 2019

View reviewed changes

arokem force-pushed the bundle_profiles branch from 9577808 to 7fff8bd Compare January 14, 2019 17:19

arokem force-pushed the bundle_profiles branch from 7dcbd28 to 011797a Compare January 26, 2019 20:23

skoudoro reviewed Feb 1, 2019

View reviewed changes

arokem force-pushed the bundle_profiles branch from 867ea80 to 8e95524 Compare February 13, 2019 14:36

arokem mentioned this pull request Feb 14, 2019

[FIX] mapmri with cvxpy 1.0.15 #1740

Merged

arokem force-pushed the bundle_profiles branch from cbf9275 to 61271d9 Compare February 15, 2019 00:11

arokem added 17 commits February 26, 2019 20:21

PEP8.

783d358

Replace some non-ascii characters in a reference.

1570f35

Replace non-ascii characters in the reference in the example as well.

4c39322

Require Streamlines object as input to the weighting function.

c700901

Require Streamlines input.

94a6786

Error handling.

a70d0ee

PEP8 in example.

3fa2f17

Add test for returning Mahalanobis instead of weights.

7d8be3a

Improve test coverage, by adding some more tests.

851dad7

Remove some rebase cruft.

07087d2

Pin the cvxpy version to 1.0.14.

bbd20a5

See: https://github.com/cvxgrp/cvxpy/issues/672

Revert "Pin the cvxpy version to 1.0.14."

6d0852d

This reverts commit cbf9275.

TST: Use extend method to generate bundles.

17985aa

Also, add references, fix typo in documentation.

Moves bundle_profile to segment.bundles and renames it.

2176cbc

Also, multiple other changes according to review by @Garyfallidis.

Adjust example accordingly.

2a949fc

PEP8 in example.

c8c2fc5

PEP8 in align.streamlinear.

4ca465a

arokem force-pushed the bundle_profiles branch from f876c9e to 4ca465a Compare February 27, 2019 04:21

arokem added 2 commits February 26, 2019 21:28

Remove outdated imports from test_streamline.

9b75a87

Add new test file.

b3a6d0e

Garyfallidis reviewed Feb 27, 2019

View reviewed changes

RF: Rename afq_tract_profile => afq_profile.

f898e1a

arokem force-pushed the bundle_profiles branch from 616abbe to f898e1a Compare February 28, 2019 03:43

Garyfallidis merged commit 666b706 into dipy:master Mar 1, 2019

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

A few tractometry functions #1695

A few tractometry functions #1695

arokem commented Dec 17, 2018

skoudoro left a comment

skoudoro Jan 3, 2019

arokem Jan 6, 2019

arokem commented Jan 4, 2019

arokem commented Jan 4, 2019

BramshQamar commented Jan 5, 2019

codecov-io commented Jan 5, 2019 •

edited

pep8speaks commented Jan 5, 2019 •

edited

arokem commented Jan 6, 2019

skoudoro left a comment

skoudoro Jan 8, 2019

skoudoro commented Jan 14, 2019

arokem commented Jan 14, 2019 via email

arokem commented Jan 22, 2019

arokem commented Jan 26, 2019

skoudoro left a comment

skoudoro Feb 1, 2019

skoudoro Feb 1, 2019

arokem commented Feb 1, 2019 via email

arokem commented Feb 12, 2019

skoudoro commented Feb 12, 2019

arokem commented Feb 13, 2019

skoudoro commented Feb 13, 2019

skoudoro commented Feb 14, 2019

arokem commented Feb 14, 2019

arokem commented Feb 14, 2019

skoudoro commented Feb 14, 2019

arokem commented Feb 27, 2019

Garyfallidis commented Feb 27, 2019

arokem commented Feb 27, 2019

Garyfallidis Feb 27, 2019

Garyfallidis commented Feb 27, 2019

Garyfallidis commented Feb 27, 2019

arokem commented Feb 27, 2019

		return w / np.sum(w, 0)


		def afq_tract_profile(data, bundle, affine=None, n_points=100,

A few tractometry functions #1695

A few tractometry functions #1695

Conversation

arokem commented Dec 17, 2018

skoudoro left a comment

Choose a reason for hiding this comment

skoudoro Jan 3, 2019

Choose a reason for hiding this comment

arokem Jan 6, 2019

Choose a reason for hiding this comment

arokem commented Jan 4, 2019

arokem commented Jan 4, 2019

BramshQamar commented Jan 5, 2019

codecov-io commented Jan 5, 2019 • edited

Codecov Report

pep8speaks commented Jan 5, 2019 • edited

Comment last updated on February 28, 2019 at 03:44 Hours UTC

arokem commented Jan 6, 2019

skoudoro left a comment

Choose a reason for hiding this comment

skoudoro Jan 8, 2019

Choose a reason for hiding this comment

skoudoro commented Jan 14, 2019

arokem commented Jan 14, 2019 via email

arokem commented Jan 22, 2019

arokem commented Jan 26, 2019

skoudoro left a comment

Choose a reason for hiding this comment

skoudoro Feb 1, 2019

Choose a reason for hiding this comment

skoudoro Feb 1, 2019

Choose a reason for hiding this comment

arokem commented Feb 1, 2019 via email

arokem commented Feb 12, 2019

skoudoro commented Feb 12, 2019

arokem commented Feb 13, 2019

skoudoro commented Feb 13, 2019

skoudoro commented Feb 14, 2019

arokem commented Feb 14, 2019

arokem commented Feb 14, 2019

skoudoro commented Feb 14, 2019

arokem commented Feb 27, 2019

Garyfallidis commented Feb 27, 2019

arokem commented Feb 27, 2019

Garyfallidis Feb 27, 2019

Choose a reason for hiding this comment

Garyfallidis commented Feb 27, 2019

Garyfallidis commented Feb 27, 2019

arokem commented Feb 27, 2019

codecov-io commented Jan 5, 2019 •

edited

pep8speaks commented Jan 5, 2019 •

edited