NF: volume computation function for nifti masks by JulianKlug · Pull Request #952 · nipy/nibabel

JulianKlug · 2020-09-09T21:58:35Z

Adds a function to compute the volume of a mask image in mm3.
Simple vanilla python equivalent to "fslstats /path/file.nii -V".

Reason: Useful for pure python projects where one can not rely on fsl.

To Do:

Add CLI

Compute volume of mask image. Equivalent to "fslstats /path/file.nii -V"

JulianKlug · 2020-09-09T22:00:11Z

A few questions to help me guide this PR:

Would this be the right location for this function? As I am still new to the source code of nibabel, I was unsure on where to add it.
Where should the CLI for this function be?

codecov · 2020-09-09T22:06:28Z

Codecov Report

Merging #952 into master will increase coverage by 0.04%.
The diff coverage is 85.29%.

@@            Coverage Diff             @@
##           master     #952      +/-   ##
==========================================
+ Coverage   91.84%   91.89%   +0.04%     
==========================================
  Files          97      101       +4     
  Lines       12428    12548     +120     
  Branches     2191     2208      +17     
==========================================
+ Hits        11415    11531     +116     
- Misses        678      680       +2     
- Partials      335      337       +2

Impacted Files	Coverage Δ
nibabel/imagestats.py	`83.33% <83.33%> (ø)`
nibabel/cmdline/stats.py	`86.36% <86.36%> (ø)`
nibabel/testing/__init__.py	`99.05% <0.00%> (-0.95%)`	⬇️
nibabel/loadsave.py	`92.59% <0.00%> (ø)`
nibabel/deprecator.py	`100.00% <0.00%> (ø)`
nibabel/cmdline/roi.py	`100.00% <0.00%> (ø)`
nibabel/conftest.py	`100.00% <0.00%> (ø)`
nibabel/casting.py	`86.32% <0.00%> (+0.17%)`	⬆️
... and 1 more

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update e65b865...35beaba. Read the comment docs.

0rC0 · 2020-09-10T11:12:17Z

I think you've forgot a return :P, something like:

return mask_volume_vx * voxel_volume_mm3

or

return mask_volume_mm3

effigies · 2020-09-10T13:06:49Z

If we want to implement something equivalent to fslstats, then I think it might be worth creating nibabel.imagestats (I think nibabel.stats might imply more analytical statistics, which are covered in nilearn) as a place to house the various functions.

The command line should be in nibabel.cmdline.<tool>, and expose the entrypoint nib-<tool>.

JulianKlug · 2020-09-10T14:40:59Z

I think you've forgot a return :P, something like:

@0rC0 actually there is a return on line 248, it is just hidden by the codecov messages.

    return mask_volume_mm3

JulianKlug · 2020-09-10T14:42:55Z

If we want to implement something equivalent to fslstats, then I think it might be worth creating nibabel.imagestats (I think nibabel.stats might imply more analytical statistics, which are covered in nilearn) as a place to house the various functions.

The command line should be in nibabel.cmdline.<tool>, and expose the entrypoint nib-<tool>.

@effigies : Sounds good to me as well. But do you think every element of fslstats should be implemented with a single release or should we implement them gradually?

effigies · 2020-09-10T14:53:37Z

I think gradually is fine. Let's call it nib-stats and start with volume calculation.

Also, don't feel compelled to mimic the fslstats flags or output conventions exactly. Clarity is a better goal.

One thing to consider: fslstats produces both voxel and mm^3 results. Do we want to follow that lead, or permit users to say --units {vox,mm3}?

pep8speaks · 2020-10-18T07:22:50Z

Hello @JulianKlug, Thank you for updating!

Cheers! There are no style issues detected in this Pull Request. 🍻 To test for issues locally, pip install flake8 and then run flake8 nibabel.

Comment last updated at 2020-10-19 21:36:20 UTC

- blank lines - variable declaration in example

JulianKlug · 2020-10-18T08:27:39Z

There seems to be an issue where the example part of the docstring is interpreted as code and not as a string which then lets the tests fail.

    >>> path = 'path/to/nifti/mask.nii'
    >>> img = nf.load(path) # path is contains a path to an example nifti mask
    >>> mask_volume(img)

Not quite sure how to resolve this...

JulianKlug · 2020-10-18T10:13:55Z

would be ready for #960 if the aforementioned issue is resolved

effigies

Thanks for pushing on this. I've left a batch of comments. Feel free to argue and ask questions. :-)

nibabel/cmdline/tests/test_stats.py

nibabel/imagestats.py

effigies · 2020-10-18T13:00:39Z

nibabel/tests/test_imagestats.py

+    assert float(vol_mm3) == 2273328656.0
+    assert float(vol_vox) == 284166082.0


$ fslstats nibabel/tests/data/anatomical.nii -V 33825 270600.000000

I'm guessing (but haven't checked) the discrepancy is from values > 1.

I thought so too, but hadn't time to check.
The volume computation is only reliable for 0-1 masks, but as I didn't find any in the sample data, I didn’t use one for the test.

nibabel/cmdline/stats.py

effigies · 2020-10-18T13:04:50Z

bin/nib-stats.py

+
+
+if __name__ == '__main__':
+    main()


These files are actually out-of-date. Just add an entry to the entry points:

nibabel/setup.cfg

Lines 71 to 81 in 917afab

[options.entry_points]

console_scripts =

nib-conform=nibabel.cmdline.conform:main

nib-ls=nibabel.cmdline.ls:main

nib-dicomfs=nibabel.cmdline.dicomfs:main

nib-diff=nibabel.cmdline.diff:main

nib-nifti-dx=nibabel.cmdline.nifti_dx:main

nib-tck2trk=nibabel.cmdline.tck2trk:main

nib-trk2tck=nibabel.cmdline.trk2tck:main

nib-roi=nibabel.cmdline.roi:main

parrec2nii=nibabel.cmdline.parrec2nii:main

Great hadn't seen that!

Let's remove this file.

effigies · 2020-10-18T13:17:57Z

nibabel/imagestats.py

+import numpy as np
+
+
+def mask_volume(img, units='mm3'):


Thinking about this a bit, if units=="vox", then the output should be integer; there's no good reason to have a float for counting non-zero voxels. That makes the types a bit messy. What if we did (simplified):

def count_nonzero_voxels(img): return np.count_nonzero(img.dataobj) def mask_volume(img): nz_vox = count_nonzero_voxels(img) return np.prod(img.header.get_zooms()[:3]) * nz_vox

The conditional logic inside this function basically evaporates, and is moved into cmdline.stats.main to select which function to call.

Sounds good to me. I have simply left the code a bit more verbose for better readability.

Co-authored-by: Chris Markiewicz <effigies@gmail.com>

…eader.pixdim Co-authored-by: Chris Markiewicz <effigies@gmail.com>

…e in mm3 - use a mask-like volume for testing

effigies

Mostly style suggestions. And please use flake8 to check for any additional issues.

effigies · 2020-10-19T02:36:27Z

bin/nib-stats.py

+
+
+if __name__ == '__main__':
+    main()


Let's remove this file.

effigies · 2020-10-19T13:46:42Z

nibabel/cmdline/tests/test_stats.py

+class Capturing(list):
+    def __enter__(self):
+        self._stdout = sys.stdout
+        sys.stdout = self._stringio = StringIO()
+        return self
+    def __exit__(self, *args):
+        self.extend(self._stringio.getvalue().splitlines())
+        del self._stringio    # free up some memory
+        sys.stdout = self._stdout


Use the capsys fixture..

thanks, didn't know about that one

effigies · 2020-10-19T13:54:36Z

nibabel/funcs.py

    """ Utility function returning True if affine is nearly diagonal """
    rzs_aff = aff[:3, :3]
    return np.allclose(rzs_aff, np.diag(np.diag(rzs_aff)))
+


We don't need to touch this file.

nibabel/imagestats.py

nibabel/tests/test_imagestats.py

Co-authored-by: Chris Markiewicz <effigies@gmail.com>

JulianKlug · 2020-10-20T05:46:11Z

Sorry for the style issues. Should be fixed now.

effigies · 2020-10-20T11:58:21Z

Thanks a lot! This is a good start to a handy utility, and I look forward to seeing it grow.

Added a volume computation function

d7c7069

Compute volume of mask image. Equivalent to "fslstats /path/file.nii -V"

effigies mentioned this pull request Oct 16, 2020

REL: 3.2.0 #960

Merged

12 tasks

Added commandline and tests for imagestats volume computation

73bff0f

Fix style errors

57f7723

- blank lines - variable declaration in example

JulianKlug marked this pull request as ready for review October 18, 2020 07:43

effigies reviewed Oct 18, 2020

View reviewed changes

JulianKlug and others added 4 commits October 18, 2020 19:43

Apply suggestions from code review: Removing unused code

e15bb8d

Co-authored-by: Chris Markiewicz <effigies@gmail.com>

Apply suggestions from code review: Use header.get_zooms instead of h…

7412ba9

…eader.pixdim Co-authored-by: Chris Markiewicz <effigies@gmail.com>

Split Mask_volume into a function for voxels and a function for volum…

d357eae

…e in mm3 - use a mask-like volume for testing

Fix failing docstring

93532de

effigies reviewed Oct 19, 2020

View reviewed changes

JulianKlug and others added 2 commits October 19, 2020 22:10

Apply suggestions from code review

2331e9d

Co-authored-by: Chris Markiewicz <effigies@gmail.com>

Style issues, use capsys

35beaba

effigies merged commit 022832e into nipy:master Oct 20, 2020

Factral mentioned this pull request Feb 27, 2023

ENH: Make imagestats module available on import nibabel #1208

Merged

		assert float(vol_mm3) == 2273328656.0
		assert float(vol_vox) == 284166082.0

	[options.entry_points]
	console_scripts =
	nib-conform=nibabel.cmdline.conform:main
	nib-ls=nibabel.cmdline.ls:main
	nib-dicomfs=nibabel.cmdline.dicomfs:main
	nib-diff=nibabel.cmdline.diff:main
	nib-nifti-dx=nibabel.cmdline.nifti_dx:main
	nib-tck2trk=nibabel.cmdline.tck2trk:main
	nib-trk2tck=nibabel.cmdline.trk2tck:main
	nib-roi=nibabel.cmdline.roi:main
	parrec2nii=nibabel.cmdline.parrec2nii:main

Conversation

JulianKlug commented Sep 9, 2020

Uh oh!

JulianKlug commented Sep 9, 2020

Uh oh!

codecov bot commented Sep 9, 2020 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codecov Report

Uh oh!

0rC0 commented Sep 10, 2020 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

effigies commented Sep 10, 2020

Uh oh!

JulianKlug commented Sep 10, 2020 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

JulianKlug commented Sep 10, 2020 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

effigies commented Sep 10, 2020

Uh oh!

pep8speaks commented Oct 18, 2020 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Comment last updated at 2020-10-19 21:36:20 UTC

Uh oh!

JulianKlug commented Oct 18, 2020 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

JulianKlug commented Oct 18, 2020

Uh oh!

effigies left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

JulianKlug Oct 18, 2020 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

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!

effigies left a comment

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!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

JulianKlug commented Oct 20, 2020

Uh oh!

effigies commented Oct 20, 2020

Uh oh!

Reviewers

Assignees

Labels

codecov bot commented Sep 9, 2020 •

edited

Loading

0rC0 commented Sep 10, 2020 •

edited

Loading

JulianKlug commented Sep 10, 2020 •

edited

Loading

JulianKlug commented Sep 10, 2020 •

edited

Loading

pep8speaks commented Oct 18, 2020 •

edited

Loading

JulianKlug commented Oct 18, 2020 •

edited

Loading

JulianKlug Oct 18, 2020 •

edited

Loading