[MRG+1] Add function for random parcellation #4955

makkostya · 2018-02-21T16:02:25Z

@agramfort @massich @mathurinm
This PR add the new function to produce random parcellation of the cortex. It grows a number (input parameter) of random not intersecting cortical regions which cover the whole surface.

Here is example of use (you should change annotation file names):

from pyface.qt import QtGui, QtCore
%matplotlib qt

import copy
import mne
from mne.datasets import sample
from surfer import Brain

subject_dir = sample.data_path() + '/subjects'
subject = 'sample'

n_parcel = 150
hemis = [0, 1]

# Parcellation
labels = mne.random_parcellation(subject, n_parcel, hemis, subject_dir, surface='white')

# Labels to annotations
annot_fname_lh = '/user/kmaksyme/home/tmp/lh.test.annot'
mne.label.write_labels_to_annot(labels,annot_fname=annot_fname_lh,hemi='lh',overwrite=True)
annot_fname_rh = '/user/kmaksyme/home/tmp/rh.test.annot'
mne.label.write_labels_to_annot(labels,annot_fname=annot_fname_rh,hemi='rh',overwrite=True)

# Visu
brain = Brain(subjects_dir=subject_dir,subject_id = subject,hemi="both",surf='inflated')
brain.add_annotation(annot_fname_lh, hemi='lh', borders=True)
brain.add_annotation(annot_fname_rh, hemi='rh', borders=True, remove_existing=False)

codecov · 2018-02-22T16:18:12Z

Codecov Report

Merging #4955 into master will increase coverage by 0.01%.
The diff coverage is 98.16%.

@@            Coverage Diff             @@
##           master    #4955      +/-   ##
==========================================
+ Coverage   87.85%   87.87%   +0.01%     
==========================================
  Files         353      353              
  Lines       62988    63177     +189     
  Branches    10783    10811      +28     
==========================================
+ Hits        55337    55515     +178     
- Misses       4909     4913       +4     
- Partials     2742     2749       +7

agramfort · 2018-02-23T17:34:25Z

mne/label.py

+        Name of the subject as in SUBJECTS_DIR.
+    n_parcel : int
+        Total number of cortical parcels
+    hemis : array | int


it should be called hemi and be a string. Following the API of pysurfer:

https://github.com/nipy/PySurfer/blob/master/surfer/viz.py#L301

I used the doc of the function "grow_labels". If need to change it, probably should do it for grow_labels too.

agramfort · 2018-02-23T17:36:09Z

mne/label.py

+    """Random cortex parcellation."""
+    labels = []
+    parcel_size = np.floor((len(vertices_[hemis[0]]) +
+                            len(vertices_[hemis[1]])) / n_parcel)


I think

len(vertices_[hemis[0]]) + len(vertices_[hemis[1]])) // n_parcel

is nicer and is the same

agramfort · 2018-02-23T17:37:00Z

mne/label.py

+
+        # prepare parcellation
+        parc = np.empty(n_vertices, dtype='int32')
+        parc[:] = -1


use np.full

https://docs.scipy.org/doc/numpy/reference/generated/numpy.full.html

agramfort · 2018-02-23T17:37:10Z

mne/label.py

+        parc[:] = -1
+
+        # initialize active sources
+        s = 0


line is useless

agramfort · 2018-02-23T17:38:18Z

mne/label.py

+            name = 'label_' + str(i)
+            label_ = Label(vertices, hemi=hemi, name=name, subject=subject)
+            labels.append(label_)
+        parc_both += [parc]


is this the same algo as in brainstorm?

I followed the same logic as brainsotm code:

define number of vertices per parcel

grow parcels till there are no free neighbor vertices or till riches max size

merge smallest parcels till rich the needed number

agramfort · 2018-02-23T17:39:59Z

mne/label.py



+def random_parcellation(subject, n_parcel, hemis, subjects_dir=None,
+                        surface='white'):


you should use np.random and have a random_state parameter to make the result reproducible.

git grep random_state

to see how we do it.

agramfort · 2018-02-23T17:45:58Z

for the lazy curious is the output of the script shared by @makkostya

larsoner · 2018-02-23T17:47:08Z

Oh the shame of having to run the example yourself to produce this output. It should be done by CircleCI and sphinx-gallery :)

agramfort · 2018-02-25T20:52:37Z

I used the doc of the function "grow_labels". If need to change it,

probably should do it for grow_labels too. hum indeed. I would recommend to use the surfer way and we'll deprecate the hemis param in grow_label in another PR. Having to pass array of int is really not pythonic

makkostya · 2018-02-25T20:57:33Z

Ok, I'll change it. And can do PR for "grow_labels" too

agramfort · 2018-02-25T21:04:42Z

thanks a lot

makkostya · 2018-02-27T12:17:45Z

hum indeed. I would recommend to use the surfer way and we'll deprecate the hemis param in grow_label in another PR.

grow_label function has parameter hemis which is different from pysurf because we can associate the hemisphere to each source seed. So may be it makes sense to keep it like it is?
But I changed it in random_parcellation.

agramfort · 2018-02-27T14:49:49Z

mne/label.py

+        parc = np.full(n_vertices, -1, dtype='int32')
+
+        # initialize active sources
+        rng = check_random_state(random_state)


I would have put the check_random_state outside of the for loop on hemi no?

agramfort · 2018-02-27T19:24:44Z

LGTM

any other review?

larsoner · 2018-02-27T20:04:25Z

LGTM but @makkostya your git / GitHub config appears to be incorrect. I should be able to click on the little icon to the left of your commits to get to your GitHub profile. It might be a matter of adding your git config email address to your GitHub account, not sure. Do you want to fix it before merge?

makkostya · 2018-02-28T09:45:01Z

@larsoner , I added my git config email address to my GitHub account, is it Ok now?

larsoner · 2018-02-28T12:41:39Z

Yep all better, thanks @makkostya

* Add functions to create random parcellation * add test for random_parcellation * add random_parcellation to python_reference * fixe pep8 * fixe pep8 mne-tools#2 * fixe pep mne-tools#3 * add random_state, small corrections * fixes random_state * FIX: Alphabetical

makkostya added 2 commits February 21, 2018 16:35

Add functions to create random parcellation

81bc7cd

add test for random_parcellation

4ced440

makkostya added 4 commits February 23, 2018 10:46

add random_parcellation to python_reference

4d8a0d7

fixe pep8

dd4d1f2

fixe pep8 mne-tools#2

eaa055a

fixe pep mne-tools#3

57e7409

agramfort reviewed Feb 23, 2018

View reviewed changes

mne/label.py Outdated

parc[:] = -1

# initialize active sources

s = 0

Copy link

Member

agramfort Feb 23, 2018

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

line is useless

agramfort reviewed Feb 23, 2018

View reviewed changes

add random_state, small corrections

78fd799

agramfort reviewed Feb 27, 2018

View reviewed changes

fixes random_state

511f5cf

agramfort changed the title ~~[WIP] Add function for random parcellation~~ [MRG+1] Add function for random parcellation Feb 27, 2018

agramfort approved these changes Feb 27, 2018

View reviewed changes

FIX: Alphabetical

160ffd8

larsoner merged commit 8ef95c9 into mne-tools:master Feb 28, 2018



		def random_parcellation(subject, n_parcel, hemis, subjects_dir=None,
		surface='white'):

Uh oh!

[MRG+1] Add function for random parcellation #4955

[MRG+1] Add function for random parcellation #4955

Uh oh!

Conversation

makkostya commented Feb 21, 2018

Uh oh!

codecov bot commented Feb 22, 2018 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codecov Report

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

agramfort commented Feb 23, 2018

Uh oh!

larsoner commented Feb 23, 2018

Uh oh!

agramfort commented Feb 25, 2018 via email

Uh oh!

makkostya commented Feb 25, 2018

Uh oh!

agramfort commented Feb 25, 2018 via email

Uh oh!

makkostya commented Feb 27, 2018

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

agramfort commented Feb 27, 2018

Uh oh!

larsoner commented Feb 27, 2018

Uh oh!

makkostya commented Feb 28, 2018

Uh oh!

larsoner commented Feb 28, 2018

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

codecov bot commented Feb 22, 2018 •

edited

Loading