Support for motion parameters produced by AFNI (FD)

[chrisgorgo]

No description provided.

[satra]

why not use the same set of support that's in ArtifactDetect?

nipype/nipype/algorithms/rapidart.py

Line 163 in 86cd8ff

parameter_source = traits.Enum("SPM", "FSL", "AFNI", "NiPy", "FSFAST",

and parameter interpretation is done here.

nipype/nipype/algorithms/rapidart.py

Line 42 in 86cd8ff

def _get_affine_matrix(params, source):

ps. the composite norm output of artifact detect is another way of measuring motion.

[codecov-io]

Codecov Report

Merging #1840 into master will decrease coverage by -0.39%.
The diff coverage is 88.88%.

@@            Coverage Diff             @@
##           master    #1840      +/-   ##
==========================================
- Coverage    73.1%   72.72%   -0.39%     
==========================================
  Files        1064     1059       -5     
  Lines       53322    52559     -763     
==========================================
- Hits        38980    38221     -759     
+ Misses      14342    14338       -4

Flag	Coverage Δ
#unittests	72.72% <88.88%> (-0.39%)	❌

Impacted Files	Coverage Δ
...lgorithms/tests/test_auto_FramewiseDisplacement.py	92.3% <ø> (-0.55%)	❌
nipype/algorithms/confounds.py	80.52% <100%> (+0.28%)	✅
nipype/algorithms/tests/test_confounds.py	93.33% <100%> (-1.41%)	❌
nipype/algorithms/rapidart.py	37.46% <100%> (+0.14%)	✅
nipype/utils/misc.py	88.97% <71.42%> (-1.03%)	❌
...faces/utility/tests/test_auto_IdentityInterface.py	69.23% <ø> (-2.2%)	❌
nipype/interfaces/tests/test_auto_S3DataGrabber.py	84.61% <ø> (-1.1%)	❌
nipype/algorithms/tests/test_auto_MergeROIs.py	84.61% <ø> (-1.1%)	❌
.../interfaces/mrtrix/tests/test_auto_MRTrixViewer.py	84.61% <ø> (-1.1%)	❌
... and 700 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 0ac8143...ff2160a. Read the comment docs.

[chrisgorgo]

All good after last commit @satra ?

[effigies]

Given that this issue occurs multiple times (see also #1379), would it be useful to have a util function that loads/normalizes motion parameter files?

[effigies]

FSFAST can be supported pretty easily, given AFNI support. See other comment. And I believe SPM and FSL parameter files are the same?

[effigies]

I'd suggest modifying the mpars rather than the diff, just to make it a little clearer what's going on. Following also adds FSFAST support.

mpars = np.loadtxt(self.inputs.in_file)  # mpars is N_t x 6 (N_t x 7 for FSFAST)
if self.inputs.parameter_source in ('AFNI', 'FSFAST'):
    mpars = mpars[np.asarray([4, 5, 3, 1, 2, 0]) + (mpars.shape[1] > 6)]
    mpars[:, 3:] *= np.pi / 180
diff = np.diff(mpars, axis=0)
diff[:, :3] *= self.inputs.radius
fd_res = np.abs(diff).sum(axis=1)

[effigies]

UGH. Do not use the above, which re-sorts the parameters. This can be done more simply by taking the last 6 columns.

mpars = np.loadtxt(self.inputs.in_file)[:, -6:]  # mpars is N_t x 6
if self.inputs.parameter_source in ('AFNI', 'FSFAST'):
    mpars[:, :3] *= np.pi / 180
diff = np.diff(mpars, axis=0)
diff[:, :3] *= self.inputs.radius
fd_res = np.abs(diff).sum(axis=1)

And this doesn't handle SPM, contra my previous comment, because SPM puts its rotation parameters first. (Though my confused comments may support the idea of solving the "load normalized motion params" issue once and reusing.)

[satra]

@effigies - if you hadn't seen my comments to chris earlier, here is the other place in nipype we deal with this:

nipype/nipype/algorithms/rapidart.py

Line 42 in 86cd8ff

def _get_affine_matrix(params, source):

[effigies]

@satra I did, and that's what I was working off of for my first comment (and completely skipped over the FSL/SPM difference). Which makes this (if we count #1379) three places where we (at least partially) solve the same problem of normalizing for parameter source. That one only handles a single row, so it's not quite a drop-in replacement, but still.

[chrisgorgo]

How does this new refactor look to you?

[effigies]

Looks reasonable.