Skip to content

Commit

Permalink
Drop unused modules, classes, and functions (#11)
Browse files Browse the repository at this point in the history 
* Remove dwi, anatomical, and fieldmap workflows.

* Remove lots of unused interfaces.

* Update __init__.py

* Remove more.

* Do more.

* Update __init__.py

* Add init_output_grid_wf back in.

* Remove QSIPrep-specific parameters and Config elements (#6)

---------

Co-authored-by: Matt Cieslak <[email protected]>

---------

Co-authored-by: Matt Cieslak <[email protected]>
  • Loading branch information
@tsalo @mattcieslak
tsalo and mattcieslak authored Aug 14, 2024
1 parent 3604b96 commit 667c5de
Show file tree
Hide file tree
Showing 64 changed files with 77 additions and 21,868 deletions.
21 changes: 2 additions & 19 deletions qsirecon/interfaces/__init__.py
View file
Original file line number Diff line number Diff line change
@@ -1,22 +1,5 @@
# -*- coding: utf-8 -*-
# emacs: -*- mode: python; py-indent-offset: 4; indent-tabs-mode: nil -*-
# vi: set ft=python sts=4 ts=4 sw=4 et:
from .bids import (
BIDSDataGrabber,
BIDSInfo,
DerivativesDataSink,
DerivativesMaybeDataSink,
)
from .confounds import DMRISummary, GatherConfounds
from .fmap import FieldToHz, FieldToRadS, Phasediff2Fieldmap, Phases2Fieldmap
from .freesurfer import (
FSDetectInputs,
FSInjectBrainExtracted,
MakeMidthickness,
RefineBrainMask,
StructuralReference,
)
from .images import Conform, ConformDwi, IntraModalMerge, ValidateImage
from .reports import AboutSummary, SubjectSummary
from .surf import NormalizeSurf
from .utils import AddTSVHeader, ConcatAffines
from .bids import DerivativesDataSink
from .images import ConformDwi, ValidateImage
235 changes: 24 additions & 211 deletions qsirecon/interfaces/anatomical.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -14,9 +14,7 @@
from pathlib import Path

import nibabel as nb
import nilearn.image as nim
import numpy as np
from dipy.segment.threshold import otsu
from nipype import logging
from nipype.interfaces.base import (
BaseInterfaceInputSpec,
Expand All @@ -27,16 +25,12 @@
traits,
)
from nipype.utils.filemanip import fname_presuffix
from pkg_resources import resource_filename as pkgr
from pkg_resources import resource_filename as pkgrf
from scipy import ndimage
from scipy.spatial import distance

from ..utils.ingress import ukb_dirname_to_bids
from .images import to_lps

LOGGER = logging.getLogger("nipype.interface")
KNOWN_TEMPLATES = ["MNI152NLin2009cAsym", "infant"]


class QSIReconAnatomicalIngressInputSpec(BaseInterfaceInputSpec):
Expand Down Expand Up @@ -198,94 +192,6 @@ def _run_interface(self, runtime):
return runtime


class _DiceOverlapInputSpec(BaseInterfaceInputSpec):
anatomical_mask = File(exists=True, mandatory=True, desc="Mask from a T1w image")
dwi_mask = File(exists=True, mandatory=True, desc="Mask from a DWI image")


class _DiceOverlapOutputSpec(TraitedSpec):
dice_score = traits.Float()


class DiceOverlap(SimpleInterface):
input_spec = _DiceOverlapInputSpec
output_spec = _DiceOverlapOutputSpec

def _run_interface(self, runtime):
t1_img = nb.load(self.inputs.anatomical_mask)
dwi_img = nb.load(self.inputs.dwi_mask)

if not t1_img.shape == dwi_img.shape:
raise Exception("Cannot compare masks with different shapes")

self._results["dice_score"] = distance.dice(
t1_img.get_fdata().flatten(), dwi_img.get_fdata().flatten()
)
return runtime


class _VoxelSizeChooserInputSpec(BaseInterfaceInputSpec):
voxel_size = traits.Float()
input_image = File(exists=True)
anisotropic_strategy = traits.Enum("min", "max", "mean", usedefault=True)


class _VoxelSizeChooserOutputSpec(TraitedSpec):
voxel_size = traits.Float()


class VoxelSizeChooser(SimpleInterface):
input_spec = _VoxelSizeChooserInputSpec
output_spec = _VoxelSizeChooserOutputSpec

def _run_interface(self, runtime):
if not isdefined(self.inputs.input_image) and not isdefined(self.inputs.voxel_size):
raise Exception("Either voxel_size or input_image need to be defined")

# A voxel size was specified without an image
if isdefined(self.inputs.voxel_size):
voxel_size = self.inputs.voxel_size
else:
# An image was provided
img = nb.load(self.inputs.input_image)
zooms = img.header.get_zooms()[:3]
if self.inputs.anisotropic_strategy == "min":
voxel_size = min(zooms)
elif self.inputs.anisotropic_strategy == "max":
voxel_size = max(zooms)
else:
voxel_size = np.round(np.mean(zooms), 2)

self._results["voxel_size"] = voxel_size
return runtime


class _FakeSegmentationInputSpec(BaseInterfaceInputSpec):
mask_file = File(exists=True, mandatory=True)


class _FakeSegmentationOutputSpec(TraitedSpec):
dseg_file = File(exists=True)


class FakeSegmentation(SimpleInterface):
input_spec = _FakeSegmentationInputSpec
output_spec = _FakeSegmentationOutputSpec

def _run_interface(self, runtime):
img = nb.load(self.inputs.mask_file)
orig_mask = img.get_fdata() > 0
eroded1 = ndimage.binary_erosion(orig_mask, iterations=3)
eroded2 = ndimage.binary_erosion(eroded1, iterations=3)
final = orig_mask.astype(int) + eroded1 + eroded2
out_img = nb.Nifti1Image(final, img.affine, header=img.header)
out_fname = fname_presuffix(self.inputs.mask_file, suffix="_dseg", newpath=runtime.cwd)
out_img.to_filename(out_fname)
self._results["dseg_file"] = out_fname

return runtime


"""
The spherical harmonic coefficients are stored as follows. First, since the
Expand Down Expand Up @@ -404,130 +310,37 @@ def calculate_order(order):
return runtime


class _DesaturateSkullInputSpec(BaseInterfaceInputSpec):
skulled_t2w_image = File(exists=True, mandatory=True, desc="Skull-on T2w image")
brain_mask_image = File(
exists=True, mandatory=True, desc="Binary brain mask in the same grid as skulled_t2w_image"
)
brain_to_skull_ratio = traits.CFloat(
8.0, usedefault=True, desc="Ratio of signal in the brain to signal in the skull"
)
class _VoxelSizeChooserInputSpec(BaseInterfaceInputSpec):
voxel_size = traits.Float()
input_image = File(exists=True)
anisotropic_strategy = traits.Enum("min", "max", "mean", usedefault=True)


class _DesaturateSkullOutputSpec(TraitedSpec):
desaturated_t2w = File(exists=True)
head_scaling_factor = traits.Float(0.0)
class _VoxelSizeChooserOutputSpec(TraitedSpec):
voxel_size = traits.Float()


class DesaturateSkull(SimpleInterface):
input_spec = _DesaturateSkullInputSpec
output_spec = _DesaturateSkullOutputSpec
class VoxelSizeChooser(SimpleInterface):
input_spec = _VoxelSizeChooserInputSpec
output_spec = _VoxelSizeChooserOutputSpec

def _run_interface(self, runtime):
if not isdefined(self.inputs.input_image) and not isdefined(self.inputs.voxel_size):
raise Exception("Either voxel_size or input_image need to be defined")

out_file = fname_presuffix(
self.inputs.skulled_t2w_image,
newpath=runtime.cwd,
suffix="_desaturated.nii",
use_ext=False,
)
skulled_img = nim.load_img(self.inputs.skulled_t2w_image)
brainmask_img = nim.load_img(self.inputs.brain_mask_image)
brain_median, nonbrain_head_median = calculate_nonbrain_saturation(
skulled_img, brainmask_img
)

actual_brain_to_skull_ratio = brain_median / nonbrain_head_median
LOGGER.info("found brain to skull ratio: %.3f", actual_brain_to_skull_ratio)
desat_data = skulled_img.get_fdata(dtype="float32").copy()
adjustment = 1.0
if actual_brain_to_skull_ratio < self.inputs.brain_to_skull_ratio:
# We need to downweight the non-brain voxels
adjustment = actual_brain_to_skull_ratio / self.inputs.brain_to_skull_ratio
LOGGER.info("Desaturating outside-brain signal by %.5f" % adjustment)
nonbrain_mask = brainmask_img.get_fdata() < 1
# Apply the adjustment
desat_data[nonbrain_mask] = desat_data[nonbrain_mask] * adjustment

desat_img = nim.new_img_like(skulled_img, desat_data, copy_header=True)
desat_img.header.set_data_dtype("float32")
desat_img.to_filename(out_file)
self._results["desaturated_t2w"] = out_file
self._results["head_scaling_factor"] = adjustment
return runtime


def calculate_nonbrain_saturation(head_img, brain_mask_img):
# Calculate the
head_data = head_img.get_fdata()
brain_mask = brain_mask_img.get_fdata() > 0

def clip_values(values):
_, top_percent = np.percentile(values, np.array([0, 99.75]), axis=None)
return np.clip(values, 0, top_percent)

nonbrain_voxels = head_data[np.logical_not(brain_mask)]
winsorized_nonbrain_voxels = clip_values(nonbrain_voxels)
threshold = otsu(winsorized_nonbrain_voxels) * 0.5

nbmask = np.zeros_like(head_img.get_fdata())
nbmask[head_data > threshold] = 2
nbmask[brain_mask] = 0

in_brain_median = np.median(head_data[brain_mask])
non_brain_head_median = np.median(head_data[nbmask > 0])

return in_brain_median, non_brain_head_median


class _GetTemplateInputSpec(BaseInterfaceInputSpec):
template_name = traits.Str("MNI152NLin2009cAsym", usedefault=True, mandatory=True)
t1_file = File(exists=True)
t2_file = File(exists=True)
mask_file = File(exists=True)
infant_mode = traits.Bool(False, usedefault=True)
anatomical_contrast = traits.Enum("T1w", "T2w", "none")


class _GetTemplateOutputSpec(BaseInterfaceInputSpec):
template_name = traits.Str()
template_file = File(exists=True)
template_brain_file = File(exists=True)
template_mask_file = File(exists=True)


class GetTemplate(SimpleInterface):
input_spec = _GetTemplateInputSpec
output_spec = _GetTemplateOutputSpec

def _run_interface(self, runtime):
self._results["template_name"] = self.inputs.template_name
contrast_name = self.inputs.anatomical_contrast.lower()
if contrast_name == "none":
LOGGER.info("Using T1w modality template for ACPC alignment")
contrast_name = "t1w"

# Cover the cases where the template images are actually in the
# qsirecon package. This is for common use cases (MNI2009cAsym and Infant)
# and legacy
if self.inputs.template_name in KNOWN_TEMPLATES or self.inputs.infant_mode:
if not self.inputs.infant_mode:
ref_img = pkgr("qsirecon", "data/mni_1mm_%s_lps.nii.gz" % contrast_name)
ref_img_brain = pkgr(
"qsirecon",
"data/mni_1mm_%s_lps_brain.nii.gz" % contrast_name,
)
ref_img_mask = pkgr("qsirecon", "data/mni_1mm_t1w_lps_brainmask.nii.gz")
else:
ref_img = pkgr("qsirecon", "data/mni_1mm_%s_lps_infant.nii.gz" % contrast_name)
ref_img_brain = pkgr(
"qsirecon", "data/mni_1mm_%s_lps_brain_infant.nii.gz" % contrast_name
)
ref_img_mask = pkgr("qsirecon", "data/mni_1mm_t1w_lps_brainmask_infant.nii.gz")
self._results["template_file"] = ref_img
self._results["template_brain_file"] = ref_img_brain
self._results["template_mask_file"] = ref_img_mask
# A voxel size was specified without an image
if isdefined(self.inputs.voxel_size):
voxel_size = self.inputs.voxel_size
else:
raise NotImplementedError("Arbitrary templates not available yet")
# An image was provided
img = nb.load(self.inputs.input_image)
zooms = img.header.get_zooms()[:3]
if self.inputs.anisotropic_strategy == "min":
voxel_size = min(zooms)
elif self.inputs.anisotropic_strategy == "max":
voxel_size = max(zooms)
else:
voxel_size = np.round(np.mean(zooms), 2)

self._results["voxel_size"] = voxel_size
return runtime
Loading

0 comments on commit 667c5de

Please sign in to comment.