Skip to content

FIX: Address issues with RobustAverage global signal measurement #607

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 1 commit into from
Jan 30, 2021
Merged

FIX: Address issues with RobustAverage global signal measurement #607

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
35 changes: 26 additions & 9 deletions niworkflows/interfaces/bold.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -17,13 +17,22 @@

class _NonsteadyStatesDetectorInputSpec(BaseInterfaceInputSpec):
in_file = File(exists=True, mandatory=True, desc="BOLD fMRI timeseries")
nonnegative = traits.Bool(True, usedefault=True,
desc="whether image voxels must be nonnegative")
n_volumes = traits.Range(
value=50,
value=40,
low=10,
high=200,
usedefault=True,
desc="drop volumes in 4D image beyond this timepoint",
)
zero_dummy_masked = traits.Range(
value=20,
low=2,
high=40,
usedefault=True,
desc="number of timepoints to average when the number of dummies is zero"
)


class _NonsteadyStatesDetectorOutputSpec(TraitedSpec):
Expand All @@ -43,8 +52,7 @@ def _run_interface(self, runtime):
img = nb.load(self.inputs.in_file)

ntotal = img.shape[-1] if img.dataobj.ndim == 4 else 1

self._results["t_mask"] = [False] * ntotal
t_mask = np.zeros((ntotal,), dtype=bool)

if ntotal == 1:
self._results["t_mask"] = [True]
Expand All @@ -53,13 +61,22 @@ def _run_interface(self, runtime):

from nipype.algorithms.confounds import is_outlier

global_signal = np.mean(
np.asanyarray(img.dataobj[..., : self.inputs.n_volumes]), axis=(0, 1, 2)
data = img.get_fdata(dtype="float32")[..., :self.inputs.n_volumes]
# Data can come with outliers showing very high numbers - preemptively prune
data = np.clip(
data,
a_min=0.0 if self.inputs.nonnegative else np.percentile(data, 0.2),
a_max=np.percentile(data, 99.8),
)
self._results["n_dummy"] = is_outlier(np.mean(data, axis=(0, 1, 2)))

start = 0
stop = self._results["n_dummy"]
if stop < 2:
stop = min(ntotal, self.inputs.n_volumes)
start = max(0, stop - self.inputs.zero_dummy_masked)

ndiscard = is_outlier(global_signal)
self._results["n_dummy"] = ndiscard
ndiscard = ndiscard or ntotal
self._results["t_mask"][:ndiscard] = [True] * ndiscard
t_mask[start:stop] = True
self._results["t_mask"] = t_mask.tolist()

return runtime
38 changes: 26 additions & 12 deletions niworkflows/interfaces/images.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -195,9 +195,9 @@ class _RobustAverageInputSpec(BaseInterfaceInputSpec):
)
t_mask = traits.List(traits.Bool, desc="List of selected timepoints to be averaged")
mc_method = traits.Enum(
None,
"AFNI",
"FSL",
None,
usedefault=True,
desc="Which software to use to perform motion correction",
)
Expand Down Expand Up @@ -257,7 +257,7 @@ def _run_interface(self, runtime):
f"Image length ({img_len} timepoints) unmatched by mask ({len(t_mask)})"
)

n_volumes = np.sum(t_mask)
n_volumes = sum(t_mask)
if n_volumes < 1:
raise ValueError("At least one volume should be selected for slicing")

Expand All @@ -267,7 +267,28 @@ def _run_interface(self, runtime):
sliced = nb.concat_images(
i for i, t in zip(nb.four_to_three(img), t_mask) if t
)
sliced.to_filename(self._results["out_volumes"])

data = sliced.get_fdata(dtype="float32")
# Data can come with outliers showing very high numbers - preemptively prune
data = np.clip(
data,
a_min=0.0 if self.inputs.nonnegative else np.percentile(data, 0.2),
a_max=np.percentile(data, 99.8),
)

gs_drift = np.mean(data, axis=(0, 1, 2))
gs_drift /= gs_drift.max()
self._results["out_drift"] = [float(i) for i in gs_drift]

data /= gs_drift
data = np.clip(
data,
a_min=0.0 if self.inputs.nonnegative else data.min(),
a_max=data.max(),
)
sliced.__class__(data, sliced.affine, sliced.header).to_filename(
self._results["out_volumes"]
)

if n_volumes == 1:
nb.squeeze_image(sliced).to_filename(self._results["out_file"])
Expand All @@ -280,11 +301,10 @@ def _run_interface(self, runtime):
res = Volreg(
in_file=self._results["out_volumes"],
args="-Fourier -twopass",
oned_matrix_save="afni-oned-matrix.xfm",
zpad=4,
outputtype="NIFTI_GZ",
).run()
self._results["out_hmc"] = res.outputs.oned_matrix_save
# self._results["out_hmc"] = res.outputs.oned_matrix_save

elif self.inputs.mc_method == "FSL":
from nipype.interfaces.fsl import MCFLIRT
Expand All @@ -297,14 +317,8 @@ def _run_interface(self, runtime):
self._results["out_hmc"] = res.outputs.mat_file

if self.inputs.mc_method:
sliced = nb.load(res.outputs.out_file)

data = np.asanyarray(sliced.dataobj)
gs_drift = np.median(data, axis=(0, 1, 2))
gs_drift /= gs_drift[0]
self._results["out_drift"] = [float(i) for i in gs_drift]
data = nb.load(res.outputs.out_file).get_fdata(dtype="float32")

data /= gs_drift[np.newaxis, np.newaxis, np.newaxis, ...]
data = np.clip(
data,
a_min=0.0 if self.inputs.nonnegative else data.min(),
Expand Down