def create_workflow(config: AttrDict, resource_pool: ResourcePool,
context: Context):
for _, rp in resource_pool[['T1w']]:
anat_image = rp[R('T1w')]
anat_deoblique = NipypeJob(interface=afni.Refit(deoblique=True),
reference='anat_deoblique')
anat_deoblique.in_file = anat_image
output_node = anat_deoblique.out_file
if config.non_local_means_filtering:
denoise = NipypeJob(interface=ants.DenoiseImage(),
reference='anat_denoise')
denoise.input_image = output_node
output_node = denoise.output_image
if config.n4_bias_field_correction:
n4 = NipypeJob(interface=ants.N4BiasFieldCorrection(
dimension=3, shrink_factor=2, copy_header=True),
reference='anat_n4')
n4.input_image = output_node
output_node = n4.output_image
anat_reorient = NipypeJob(interface=afni.Resample(
orientation='RPI', outputtype='NIFTI_GZ'),
reference='anat_reorient')
anat_reorient.in_file = output_node
rp[R('T1w', label='reorient')] = anat_reorient.out_file
def non_local_means(wf, cfg, strat_pool, pipe_num, opt=None):
'''
{"name": "nlm_filtering",
"config": ["anatomical_preproc"],
"switch": ["non_local_means_filtering"],
"option_key": "None",
"option_val": "None",
"inputs": [["desc-preproc_T1w", "desc-reorient_T1w", "T1w"]],
"outputs": ["desc-preproc_T1w"]}
'''
denoise = pe.Node(interface=ants.DenoiseImage(),
name=f'anat_denoise_{pipe_num}')
node, out = strat_pool.get_data(
['desc-preproc_T1w', 'desc-reorient_T1w', 'T1w'])
wf.connect(node, out, denoise, 'input_image')
outputs = {'desc-preproc_T1w': (denoise, 'output_image')}
return (wf, outputs)
def create_anat_preproc(template_path=None,
mask_path=None,
regmask_path=None,
method='afni',
already_skullstripped=False,
non_local_means_filtering=True,
n4_correction=True,
wf_name='anat_preproc'):
"""
The main purpose of this workflow is to process T1 scans. Raw mprage file is deobliqued, reoriented
into RPI and skullstripped. Also, a whole brain only mask is generated from the skull stripped image
for later use in registration.
Returns
-------
anat_preproc : workflow
Anatomical Preprocessing Workflow
Notes
-----
`Source <https://github.com/FCP-INDI/C-PAC/blob/master/CPAC/anat_preproc/anat_preproc.py>`_
Workflow Inputs::
inputspec.anat : string
User input anatomical (T1) Image, in any of the 8 orientations
Workflow Outputs::
outputspec.refit : string
Path to deobliqued anatomical image
outputspec.reorient : string
Path to RPI oriented anatomical image
outputspec.skullstrip : string
Path to skull stripped RPI oriented mprage file with normalized intensities.
outputspec.brain : string
Path to skull stripped RPI brain image with original intensity values and not normalized or scaled.
Order of commands:
- Deobliqing the scans. ::
3drefit -deoblique mprage.nii.gz
- Re-orienting the Image into Right-to-Left Posterior-to-Anterior Inferior-to-Superior (RPI) orientation ::
3dresample -orient RPI
-prefix mprage_RPI.nii.gz
-inset mprage.nii.gz
- Skull-Stripping the image ::
Using AFNI ::
3dSkullStrip -input mprage_RPI.nii.gz
-o_ply mprage_RPI_3dT.nii.gz
or using BET ::
bet mprage_RPI.nii.gz
- The skull-stripping step modifies the intensity values. To get back the original intensity values, we do an element wise product of RPI data with step function of skull-stripped data ::
3dcalc -a mprage_RPI.nii.gz
-b mprage_RPI_3dT.nii.gz
-expr 'a*step(b)'
-prefix mprage_RPI_3dc.nii.gz
High Level Workflow Graph:
.. image:: ../images/anatpreproc_graph.dot.png
:width: 500
Detailed Workflow Graph:
.. image:: ../images/anatpreproc_graph_detailed.dot.png
:width: 500
Examples
--------
>>> from CPAC.anat_preproc import create_anat_preproc
>>> preproc = create_anat_preproc()
>>> preproc.inputs.inputspec.anat = 'sub1/anat/mprage.nii.gz'
>>> preproc.run() #doctest: +SKIP
"""
preproc = pe.Workflow(name=wf_name)
inputnode = pe.Node(util.IdentityInterface(fields=['anat', 'brain_mask']),
name='inputspec')
outputnode = pe.Node(util.IdentityInterface(
fields=['refit', 'reorient', 'skullstrip', 'brain', 'brain_mask']),
name='outputspec')
anat_deoblique = pe.Node(interface=afni.Refit(), name='anat_deoblique')
anat_deoblique.inputs.deoblique = True
preproc.connect(inputnode, 'anat', anat_deoblique, 'in_file')
preproc.connect(anat_deoblique, 'out_file', outputnode, 'refit')
# Disable non_local_means_filtering and n4_correction when run niworkflows-ants
if method == 'niworkflows-ants':
non_local_means_filtering = False
n4_correction = False
if non_local_means_filtering and n4_correction:
denoise = pe.Node(interface=ants.DenoiseImage(), name='anat_denoise')
preproc.connect(anat_deoblique, 'out_file', denoise, 'input_image')
n4 = pe.Node(interface=ants.N4BiasFieldCorrection(dimension=3,
shrink_factor=2,
copy_header=True),
name='anat_n4')
preproc.connect(denoise, 'output_image', n4, 'input_image')
elif non_local_means_filtering and not n4_correction:
denoise = pe.Node(interface=ants.DenoiseImage(), name='anat_denoise')
preproc.connect(anat_deoblique, 'out_file', denoise, 'input_image')
elif not non_local_means_filtering and n4_correction:
n4 = pe.Node(interface=ants.N4BiasFieldCorrection(dimension=3,
shrink_factor=2,
copy_header=True),
name='anat_n4')
preproc.connect(anat_deoblique, 'out_file', n4, 'input_image')
# Anatomical reorientation
anat_reorient = pe.Node(interface=afni.Resample(), name='anat_reorient')
anat_reorient.inputs.orientation = 'RPI'
anat_reorient.inputs.outputtype = 'NIFTI_GZ'
if n4_correction:
preproc.connect(n4, 'output_image', anat_reorient, 'in_file')
elif non_local_means_filtering and not n4_correction:
preproc.connect(denoise, 'output_image', anat_reorient, 'in_file')
else:
preproc.connect(anat_deoblique, 'out_file', anat_reorient, 'in_file')
preproc.connect(anat_reorient, 'out_file', outputnode, 'reorient')
if already_skullstripped:
anat_skullstrip = pe.Node(
interface=util.IdentityInterface(fields=['out_file']),
name='anat_skullstrip')
preproc.connect(anat_reorient, 'out_file', anat_skullstrip, 'out_file')
preproc.connect(anat_skullstrip, 'out_file', outputnode, 'skullstrip')
preproc.connect(anat_skullstrip, 'out_file', outputnode, 'brain')
else:
if method == 'afni':
# Skull-stripping using AFNI 3dSkullStrip
inputnode_afni = pe.Node(util.IdentityInterface(fields=[
'shrink_factor', 'var_shrink_fac', 'shrink_fac_bot_lim',
'avoid_vent', 'niter', 'pushout', 'touchup', 'fill_hole',
'avoid_eyes', 'use_edge', 'exp_frac', 'smooth_final',
'push_to_edge', 'use_skull', 'perc_int', 'max_inter_iter',
'blur_fwhm', 'fac', 'monkey'
]),
name='AFNI_options')
skullstrip_args = pe.Node(util.Function(
input_names=[
'spat_norm', 'spat_norm_dxyz', 'shrink_fac',
'var_shrink_fac', 'shrink_fac_bot_lim', 'avoid_vent',
'niter', 'pushout', 'touchup', 'fill_hole', 'avoid_eyes',
'use_edge', 'exp_frac', 'smooth_final', 'push_to_edge',
'use_skull', 'perc_int', 'max_inter_iter', 'blur_fwhm',
'fac', 'monkey'
],
output_names=['expr'],
function=create_3dskullstrip_arg_string),
name='anat_skullstrip_args')
preproc.connect([(inputnode_afni, skullstrip_args,
[('shrink_factor', 'shrink_fac'),
('var_shrink_fac', 'var_shrink_fac'),
('shrink_fac_bot_lim', 'shrink_fac_bot_lim'),
('avoid_vent', 'avoid_vent'),
('niter', 'niter'), ('pushout', 'pushout'),
('touchup', 'touchup'),
('fill_hole', 'fill_hole'),
('avoid_eyes', 'avoid_eyes'),
('use_edge', 'use_edge'),
('exp_frac', 'exp_frac'),
('smooth_final', 'smooth_final'),
('push_to_edge', 'push_to_edge'),
('use_skull', 'use_skull'),
('perc_int', 'perc_int'),
('max_inter_iter', 'max_inter_iter'),
('blur_fwhm', 'blur_fwhm'), ('fac', 'fac'),
('monkey', 'monkey')])])
anat_skullstrip = pe.Node(interface=afni.SkullStrip(),
name='anat_skullstrip')
anat_skullstrip.inputs.outputtype = 'NIFTI_GZ'
preproc.connect(anat_reorient, 'out_file', anat_skullstrip,
'in_file')
preproc.connect(skullstrip_args, 'expr', anat_skullstrip, 'args')
preproc.connect(anat_skullstrip, 'out_file', outputnode,
'skullstrip')
# Apply skull-stripping step mask to original volume
anat_skullstrip_orig_vol = pe.Node(interface=afni.Calc(),
name='anat_skullstrip_orig_vol')
anat_skullstrip_orig_vol.inputs.expr = 'a*step(b)'
anat_skullstrip_orig_vol.inputs.outputtype = 'NIFTI_GZ'
preproc.connect(anat_reorient, 'out_file',
anat_skullstrip_orig_vol, 'in_file_a')
anat_brain_mask = pe.Node(interface=afni.Calc(),
name='anat_brain_mask')
anat_brain_mask.inputs.expr = 'step(a)'
anat_brain_mask.inputs.outputtype = 'NIFTI_GZ'
preproc.connect(anat_skullstrip, 'out_file', anat_brain_mask,
'in_file_a')
preproc.connect(anat_skullstrip, 'out_file',
anat_skullstrip_orig_vol, 'in_file_b')
preproc.connect(anat_brain_mask, 'out_file', outputnode,
'brain_mask')
preproc.connect(anat_skullstrip_orig_vol, 'out_file', outputnode,
'brain')
elif method == 'fsl':
# Skull-stripping using FSL BET
inputnode_bet = pe.Node(util.IdentityInterface(fields=[
'frac', 'mask_boolean', 'mesh_boolean', 'outline', 'padding',
'radius', 'reduce_bias', 'remove_eyes', 'robust', 'skull',
'surfaces', 'threshold', 'vertical_gradient'
]),
name='BET_options')
anat_skullstrip = pe.Node(interface=fsl.BET(),
name='anat_skullstrip')
preproc.connect(anat_reorient, 'out_file', anat_skullstrip,
'in_file')
preproc.connect([(inputnode_bet, anat_skullstrip, [
('frac', 'frac'),
('mask_boolean', 'mask'),
('mesh_boolean', 'mesh'),
('outline', 'outline'),
('padding', 'padding'),
('radius', 'radius'),
('reduce_bias', 'reduce_bias'),
('remove_eyes', 'remove_eyes'),
('robust', 'robust'),
('skull', 'skull'),
('surfaces', 'surfaces'),
('threshold', 'threshold'),
('vertical_gradient', 'vertical_gradient'),
])])
preproc.connect(anat_skullstrip, 'out_file', outputnode,
'skullstrip')
# Apply skull-stripping step mask to original volume
anat_skullstrip_orig_vol = pe.Node(interface=afni.Calc(),
name='anat_skullstrip_orig_vol')
anat_skullstrip_orig_vol.inputs.expr = 'a*step(b)'
anat_skullstrip_orig_vol.inputs.outputtype = 'NIFTI_GZ'
preproc.connect(anat_reorient, 'out_file',
anat_skullstrip_orig_vol, 'in_file_a')
preproc.connect(anat_skullstrip, 'out_file',
anat_skullstrip_orig_vol, 'in_file_b')
preproc.connect(anat_skullstrip, 'mask_file', outputnode,
'brain_mask')
preproc.connect(anat_skullstrip_orig_vol, 'out_file', outputnode,
'brain')
elif method == 'niworkflows-ants':
# Skull-stripping using niworkflows-ants
anat_skullstrip_ants = init_brain_extraction_wf(
tpl_target_path=template_path,
tpl_mask_path=mask_path,
tpl_regmask_path=regmask_path,
name='anat_skullstrip_ants')
preproc.connect(anat_reorient, 'out_file', anat_skullstrip_ants,
'inputnode.in_files')
preproc.connect(anat_skullstrip_ants, 'copy_xform.out_file',
outputnode, 'skullstrip')
preproc.connect(anat_skullstrip_ants, 'copy_xform.out_file',
outputnode, 'brain')
preproc.connect(anat_skullstrip_ants,
'atropos_wf.copy_xform.out_mask', outputnode,
'brain_mask')
elif method == 'mask':
anat_skullstrip_orig_vol = pe.Node(interface=afni.Calc(),
name='anat_skullstrip_orig_vol')
anat_skullstrip_orig_vol.inputs.expr = 'a*step(b)'
anat_skullstrip_orig_vol.inputs.outputtype = 'NIFTI_GZ'
preproc.connect(anat_reorient, 'out_file',
anat_skullstrip_orig_vol, 'in_file_a')
preproc.connect(inputnode, 'brain_mask', anat_skullstrip_orig_vol,
'in_file_b')
preproc.connect(inputnode, 'brain_mask', outputnode, 'brain_mask')
preproc.connect(anat_skullstrip_orig_vol, 'out_file', outputnode,
'brain')
return preproc
def create_anat_preproc(method='afni',
already_skullstripped=False,
c=None,
wf_name='anat_preproc'):
"""The main purpose of this workflow is to process T1 scans. Raw mprage file is deobliqued, reoriented
into RPI and skullstripped. Also, a whole brain only mask is generated from the skull stripped image
for later use in registration.
Returns
-------
anat_preproc : workflow
Anatomical Preprocessing Workflow
Notes
-----
`Source <https://github.com/FCP-INDI/C-PAC/blob/master/CPAC/anat_preproc/anat_preproc.py>`_
Workflow Inputs::
inputspec.anat : string
User input anatomical (T1) Image, in any of the 8 orientations
Workflow Outputs::
outputspec.refit : string
Path to deobliqued anatomical image
outputspec.reorient : string
Path to RPI oriented anatomical image
outputspec.skullstrip : string
Path to skull stripped RPI oriented mprage file with normalized intensities.
outputspec.brain : string
Path to skull stripped RPI brain image with original intensity values and not normalized or scaled.
Order of commands:
- Deobliqing the scans. ::
3drefit -deoblique mprage.nii.gz
- Re-orienting the Image into Right-to-Left Posterior-to-Anterior Inferior-to-Superior (RPI) orientation ::
3dresample -orient RPI
-prefix mprage_RPI.nii.gz
-inset mprage.nii.gz
- Skull-Stripping the image ::
Using AFNI ::
3dSkullStrip -input mprage_RPI.nii.gz
-o_ply mprage_RPI_3dT.nii.gz
or using BET ::
bet mprage_RPI.nii.gz
- The skull-stripping step modifies the intensity values. To get back the original intensity values, we do an element wise product of RPI data with step function of skull-stripped data ::
3dcalc -a mprage_RPI.nii.gz
-b mprage_RPI_3dT.nii.gz
-expr 'a*step(b)'
-prefix mprage_RPI_3dc.nii.gz
High Level Workflow Graph:
.. image:: ../images/anatpreproc_graph.dot.png
:width: 500
Detailed Workflow Graph:
.. image:: ../images/anatpreproc_graph_detailed.dot.png
:width: 500
Examples
--------
>>> from CPAC.anat_preproc import create_anat_preproc
>>> preproc = create_anat_preproc()
>>> preproc.inputs.inputspec.anat = 'sub1/anat/mprage.nii.gz'
>>> preproc.run() #doctest: +SKIP
"""
preproc = pe.Workflow(name=wf_name)
inputnode = pe.Node(util.IdentityInterface(fields=['anat', 'brain_mask']),
name='inputspec')
outputnode = pe.Node(util.IdentityInterface(
fields=['refit', 'reorient', 'skullstrip', 'brain', 'brain_mask']),
name='outputspec')
anat_deoblique = pe.Node(interface=afni.Refit(), name='anat_deoblique')
anat_deoblique.inputs.deoblique = True
preproc.connect(inputnode, 'anat', anat_deoblique, 'in_file')
preproc.connect(anat_deoblique, 'out_file', outputnode, 'refit')
# Disable non_local_means_filtering and n4_bias_field_correction when run niworkflows-ants
if method == 'niworkflows-ants':
c.non_local_means_filtering = False
c.n4_bias_field_correction = False
if c.non_local_means_filtering and c.n4_bias_field_correction:
denoise = pe.Node(interface=ants.DenoiseImage(), name='anat_denoise')
preproc.connect(anat_deoblique, 'out_file', denoise, 'input_image')
n4 = pe.Node(interface=ants.N4BiasFieldCorrection(dimension=3,
shrink_factor=2,
copy_header=True),
name='anat_n4')
preproc.connect(denoise, 'output_image', n4, 'input_image')
elif c.non_local_means_filtering and not c.n4_bias_field_correction:
denoise = pe.Node(interface=ants.DenoiseImage(), name='anat_denoise')
preproc.connect(anat_deoblique, 'out_file', denoise, 'input_image')
elif not c.non_local_means_filtering and c.n4_bias_field_correction:
n4 = pe.Node(interface=ants.N4BiasFieldCorrection(dimension=3,
shrink_factor=2,
copy_header=True),
name='anat_n4')
preproc.connect(anat_deoblique, 'out_file', n4, 'input_image')
# Anatomical reorientation
anat_reorient = pe.Node(interface=afni.Resample(), name='anat_reorient')
anat_reorient.inputs.orientation = 'RPI'
anat_reorient.inputs.outputtype = 'NIFTI_GZ'
if c.n4_bias_field_correction:
preproc.connect(n4, 'output_image', anat_reorient, 'in_file')
elif c.non_local_means_filtering and not c.n4_bias_field_correction:
preproc.connect(denoise, 'output_image', anat_reorient, 'in_file')
else:
preproc.connect(anat_deoblique, 'out_file', anat_reorient, 'in_file')
preproc.connect(anat_reorient, 'out_file', outputnode, 'reorient')
if already_skullstripped:
anat_skullstrip = pe.Node(
interface=util.IdentityInterface(fields=['out_file']),
name='anat_skullstrip')
preproc.connect(anat_reorient, 'out_file', anat_skullstrip, 'out_file')
preproc.connect(anat_skullstrip, 'out_file', outputnode, 'skullstrip')
preproc.connect(anat_skullstrip, 'out_file', outputnode, 'brain')
else:
if method == 'afni':
# Skull-stripping using AFNI 3dSkullStrip
inputnode_afni = pe.Node(util.IdentityInterface(fields=[
'mask_vol', 'shrink_factor', 'var_shrink_fac',
'shrink_fac_bot_lim', 'avoid_vent', 'niter', 'pushout',
'touchup', 'fill_hole', 'avoid_eyes', 'use_edge', 'exp_frac',
'smooth_final', 'push_to_edge', 'use_skull', 'perc_int',
'max_inter_iter', 'blur_fwhm', 'fac', 'monkey'
]),
name='AFNI_options')
skullstrip_args = pe.Node(util.Function(
input_names=[
'spat_norm', 'spat_norm_dxyz', 'mask_vol', 'shrink_fac',
'var_shrink_fac', 'shrink_fac_bot_lim', 'avoid_vent',
'niter', 'pushout', 'touchup', 'fill_hole', 'avoid_eyes',
'use_edge', 'exp_frac', 'smooth_final', 'push_to_edge',
'use_skull', 'perc_int', 'max_inter_iter', 'blur_fwhm',
'fac', 'monkey'
],
output_names=['expr'],
function=create_3dskullstrip_arg_string),
name='anat_skullstrip_args')
preproc.connect([(inputnode_afni, skullstrip_args,
[('mask_vol', 'mask_vol'),
('shrink_factor', 'shrink_fac'),
('var_shrink_fac', 'var_shrink_fac'),
('shrink_fac_bot_lim', 'shrink_fac_bot_lim'),
('avoid_vent', 'avoid_vent'),
('niter', 'niter'), ('pushout', 'pushout'),
('touchup', 'touchup'),
('fill_hole', 'fill_hole'),
('avoid_eyes', 'avoid_eyes'),
('use_edge', 'use_edge'),
('exp_frac', 'exp_frac'),
('smooth_final', 'smooth_final'),
('push_to_edge', 'push_to_edge'),
('use_skull', 'use_skull'),
('perc_int', 'perc_int'),
('max_inter_iter', 'max_inter_iter'),
('blur_fwhm', 'blur_fwhm'), ('fac', 'fac'),
('monkey', 'monkey')])])
anat_skullstrip = pe.Node(interface=afni.SkullStrip(),
name='anat_skullstrip')
anat_skullstrip.inputs.outputtype = 'NIFTI_GZ'
preproc.connect(anat_reorient, 'out_file', anat_skullstrip,
'in_file')
preproc.connect(skullstrip_args, 'expr', anat_skullstrip, 'args')
# Generate anatomical brain mask
anat_brain_mask = pe.Node(interface=afni.Calc(),
name='anat_brain_mask')
anat_brain_mask.inputs.expr = 'step(a)'
anat_brain_mask.inputs.outputtype = 'NIFTI_GZ'
preproc.connect(anat_skullstrip, 'out_file', anat_brain_mask,
'in_file_a')
# Apply skull-stripping step mask to original volume
anat_skullstrip_orig_vol = pe.Node(interface=afni.Calc(),
name='anat_skullstrip_orig_vol')
anat_skullstrip_orig_vol.inputs.expr = 'a*step(b)'
anat_skullstrip_orig_vol.inputs.outputtype = 'NIFTI_GZ'
preproc.connect(anat_reorient, 'out_file',
anat_skullstrip_orig_vol, 'in_file_a')
preproc.connect(anat_brain_mask, 'out_file',
anat_skullstrip_orig_vol, 'in_file_b')
preproc.connect(anat_brain_mask, 'out_file', outputnode,
'brain_mask')
preproc.connect(anat_skullstrip_orig_vol, 'out_file', outputnode,
'brain')
elif method == 'fsl':
# Skull-stripping using FSL BET
inputnode_bet = pe.Node(util.IdentityInterface(fields=[
'frac', 'mask_boolean', 'mesh_boolean', 'outline', 'padding',
'radius', 'reduce_bias', 'remove_eyes', 'robust', 'skull',
'surfaces', 'threshold', 'vertical_gradient'
]),
name='BET_options')
anat_skullstrip = pe.Node(interface=fsl.BET(),
name='anat_skullstrip')
anat_skullstrip.inputs.output_type = 'NIFTI_GZ'
preproc.connect(anat_reorient, 'out_file', anat_skullstrip,
'in_file')
preproc.connect([(inputnode_bet, anat_skullstrip, [
('frac', 'frac'),
('mask_boolean', 'mask'),
('mesh_boolean', 'mesh'),
('outline', 'outline'),
('padding', 'padding'),
('radius', 'radius'),
('reduce_bias', 'reduce_bias'),
('remove_eyes', 'remove_eyes'),
('robust', 'robust'),
('skull', 'skull'),
('surfaces', 'surfaces'),
('threshold', 'threshold'),
('vertical_gradient', 'vertical_gradient'),
])])
preproc.connect(anat_skullstrip, 'out_file', outputnode,
'skullstrip')
# Apply skull-stripping step mask to original volume
anat_skullstrip_orig_vol = pe.Node(interface=afni.Calc(),
name='anat_skullstrip_orig_vol')
anat_skullstrip_orig_vol.inputs.expr = 'a*step(b)'
anat_skullstrip_orig_vol.inputs.outputtype = 'NIFTI_GZ'
preproc.connect(anat_reorient, 'out_file',
anat_skullstrip_orig_vol, 'in_file_a')
preproc.connect(anat_skullstrip, 'out_file',
anat_skullstrip_orig_vol, 'in_file_b')
preproc.connect(anat_skullstrip, 'mask_file', outputnode,
'brain_mask')
preproc.connect(anat_skullstrip_orig_vol, 'out_file', outputnode,
'brain')
elif method == 'niworkflows-ants':
# Skull-stripping using niworkflows-ants
anat_skullstrip_ants = init_brain_extraction_wf(
tpl_target_path=c.niworkflows_ants_template_path,
tpl_mask_path=c.niworkflows_ants_mask_path,
tpl_regmask_path=c.niworkflows_ants_regmask_path,
name='anat_skullstrip_ants')
preproc.connect(anat_reorient, 'out_file', anat_skullstrip_ants,
'inputnode.in_files')
preproc.connect(anat_skullstrip_ants, 'copy_xform.out_file',
outputnode, 'skullstrip')
preproc.connect(anat_skullstrip_ants, 'copy_xform.out_file',
outputnode, 'brain')
preproc.connect(anat_skullstrip_ants,
'atropos_wf.copy_xform.out_mask', outputnode,
'brain_mask')
elif method == 'mask':
brain_mask_deoblique = pe.Node(interface=afni.Refit(),
name='brain_mask_deoblique')
brain_mask_deoblique.inputs.deoblique = True
preproc.connect(inputnode, 'brain_mask', brain_mask_deoblique,
'in_file')
brain_mask_reorient = pe.Node(interface=afni.Resample(),
name='brain_mask_reorient')
brain_mask_reorient.inputs.orientation = 'RPI'
brain_mask_reorient.inputs.outputtype = 'NIFTI_GZ'
preproc.connect(brain_mask_deoblique, 'out_file',
brain_mask_reorient, 'in_file')
anat_skullstrip_orig_vol = pe.Node(interface=afni.Calc(),
name='anat_skullstrip_orig_vol')
anat_skullstrip_orig_vol.inputs.expr = 'a*step(b)'
anat_skullstrip_orig_vol.inputs.outputtype = 'NIFTI_GZ'
preproc.connect(anat_reorient, 'out_file',
anat_skullstrip_orig_vol, 'in_file_a')
preproc.connect(brain_mask_reorient, 'out_file',
anat_skullstrip_orig_vol, 'in_file_b')
preproc.connect(brain_mask_reorient, 'out_file', outputnode,
'brain_mask')
preproc.connect(anat_skullstrip_orig_vol, 'out_file', outputnode,
'brain')
elif method == 'unet':
"""
UNet
options (following numbers are default):
input_slice: 3
conv_block: 5
kernel_root: 16
rescale_dim: 256
"""
# TODO: add options to pipeline_config
train_model = UNet2d(dim_in=3, num_conv_block=5, kernel_root=16)
unet_path = check_for_s3(c.unet_model)
checkpoint = torch.load(unet_path, map_location={'cuda:0': 'cpu'})
train_model.load_state_dict(checkpoint['state_dict'])
model = nn.Sequential(train_model, nn.Softmax2d())
# create a node called unet_mask
unet_mask = pe.Node(util.Function(input_names=['model', 'cimg_in'],
output_names=['out_path'],
function=predict_volumes),
name='unet_mask')
unet_mask.inputs.model = model
preproc.connect(anat_reorient, 'out_file', unet_mask, 'cimg_in')
"""
Revised mask with ANTs
"""
# fslmaths <whole head> -mul <mask> brain.nii.gz
unet_masked_brain = pe.Node(interface=fsl.MultiImageMaths(),
name='unet_masked_brain')
unet_masked_brain.inputs.op_string = "-mul %s"
preproc.connect(anat_reorient, 'out_file', unet_masked_brain,
'in_file')
preproc.connect(unet_mask, 'out_path', unet_masked_brain,
'operand_files')
# flirt -v -dof 6 -in brain.nii.gz -ref NMT_SS_0.5mm.nii.gz -o brain_rot2atl -omat brain_rot2atl.mat -interp sinc
# TODO change it to ANTs linear transform
native_brain_to_template_brain = pe.Node(
interface=fsl.FLIRT(), name='native_brain_to_template_brain')
native_brain_to_template_brain.inputs.reference = c.template_brain_only_for_anat
native_brain_to_template_brain.inputs.dof = 6
native_brain_to_template_brain.inputs.interp = 'sinc'
preproc.connect(unet_masked_brain, 'out_file',
native_brain_to_template_brain, 'in_file')
# flirt -in head.nii.gz -ref NMT_0.5mm.nii.gz -o head_rot2atl -applyxfm -init brain_rot2atl.mat
# TODO change it to ANTs linear transform
native_head_to_template_head = pe.Node(
interface=fsl.FLIRT(), name='native_head_to_template_head')
native_head_to_template_head.inputs.reference = c.template_skull_for_anat
native_head_to_template_head.inputs.apply_xfm = True
preproc.connect(anat_reorient, 'out_file',
native_head_to_template_head, 'in_file')
preproc.connect(native_brain_to_template_brain, 'out_matrix_file',
native_head_to_template_head, 'in_matrix_file')
# fslmaths NMT_SS_0.5mm.nii.gz -bin templateMask.nii.gz
template_brain_mask = pe.Node(interface=fsl.maths.MathsCommand(),
name='template_brain_mask')
template_brain_mask.inputs.in_file = c.template_brain_only_for_anat
template_brain_mask.inputs.args = '-bin'
# ANTS 3 -m CC[head_rot2atl.nii.gz,NMT_0.5mm.nii.gz,1,5] -t SyN[0.25] -r Gauss[3,0] -o atl2T1rot -i 60x50x20 --use-Histogram-Matching --number-of-affine-iterations 10000x10000x10000x10000x10000 --MI-option 32x16000
ants_template_head_to_template = pe.Node(
interface=ants.Registration(),
name='template_head_to_template')
ants_template_head_to_template.inputs.metric = ['CC']
ants_template_head_to_template.inputs.metric_weight = [1, 5]
ants_template_head_to_template.inputs.moving_image = c.template_skull_for_anat
ants_template_head_to_template.inputs.transforms = ['SyN']
ants_template_head_to_template.inputs.transform_parameters = [
(0.25, )
]
ants_template_head_to_template.inputs.interpolation = 'NearestNeighbor'
ants_template_head_to_template.inputs.number_of_iterations = [[
60, 50, 20
]]
ants_template_head_to_template.inputs.smoothing_sigmas = [[
0.6, 0.2, 0.0
]]
ants_template_head_to_template.inputs.shrink_factors = [[4, 2, 1]]
ants_template_head_to_template.inputs.convergence_threshold = [
1.e-8
]
preproc.connect(native_head_to_template_head, 'out_file',
ants_template_head_to_template, 'fixed_image')
# antsApplyTransforms -d 3 -i templateMask.nii.gz -t atl2T1rotWarp.nii.gz atl2T1rotAffine.txt -r brain_rot2atl.nii.gz -o brain_rot2atl_mask.nii.gz
template_head_transform_to_template = pe.Node(
interface=ants.ApplyTransforms(),
name='template_head_transform_to_template')
template_head_transform_to_template.inputs.dimension = 3
preproc.connect(template_brain_mask, 'out_file',
template_head_transform_to_template, 'input_image')
preproc.connect(native_brain_to_template_brain, 'out_file',
template_head_transform_to_template,
'reference_image')
preproc.connect(ants_template_head_to_template,
'forward_transforms',
template_head_transform_to_template, 'transforms')
# convert_xfm -omat brain_rot2native.mat -inverse brain_rot2atl.mat
invt = pe.Node(interface=fsl.ConvertXFM(), name='convert_xfm')
invt.inputs.invert_xfm = True
preproc.connect(native_brain_to_template_brain, 'out_matrix_file',
invt, 'in_file')
# flirt -in brain_rot2atl_mask.nii.gz -ref brain.nii.gz -o brain_mask.nii.gz -applyxfm -init brain_rot2native.mat
template_brain_to_native_brain = pe.Node(
interface=fsl.FLIRT(), name='template_brain_to_native_brain')
template_brain_to_native_brain.inputs.apply_xfm = True
preproc.connect(template_head_transform_to_template,
'output_image', template_brain_to_native_brain,
'in_file')
preproc.connect(unet_masked_brain, 'out_file',
template_brain_to_native_brain, 'reference')
preproc.connect(invt, 'out_file', template_brain_to_native_brain,
'in_matrix_file')
# fslmaths brain_mask.nii.gz -thr .5 -bin brain_mask_thr.nii.gz
refined_mask = pe.Node(interface=fsl.Threshold(),
name='refined_mask')
refined_mask.inputs.thresh = 0.5
preproc.connect(template_brain_to_native_brain, 'out_file',
refined_mask, 'in_file')
# get a new brain with mask
refined_brain = pe.Node(interface=fsl.MultiImageMaths(),
name='refined_brain')
refined_brain.inputs.op_string = "-mul %s"
preproc.connect(anat_reorient, 'out_file', refined_brain,
'in_file')
preproc.connect(refined_mask, 'out_file', refined_brain,
'operand_files')
preproc.connect(refined_mask, 'out_file', outputnode, 'brain_mask')
preproc.connect(refined_brain, 'out_file', outputnode, 'brain')
return preproc
def create_anat_preproc(method='afni', already_skullstripped=False,
config=None, acpc_target='whole-head', wf_name='anat_preproc'):
"""The main purpose of this workflow is to process T1 scans. Raw mprage
file is deobliqued, reoriented into RPI and skullstripped. Also, a whole
brain only mask is generated from the skull stripped image for later use
in registration.
Returns
-------
anat_preproc : workflow
Anatomical Preprocessing Workflow
Notes
-----
`Source <https://github.com/FCP-INDI/C-PAC/blob/master/CPAC/anat_preproc/anat_preproc.py>`_
Workflow Inputs::
inputspec.anat : string
User input anatomical (T1) Image, in any of the 8 orientations
Workflow Outputs::
outputspec.refit : string
Path to deobliqued anatomical image
outputspec.reorient : string
Path to RPI oriented anatomical image
outputspec.skullstrip : string
Path to skull stripped RPI oriented mprage file with normalized
intensities.
outputspec.brain : string
Path to skull stripped RPI brain image with original intensity
values and not normalized or scaled.
Order of commands:
- Deobliqing the scans. ::
3drefit -deoblique mprage.nii.gz
- Re-orienting the Image into Right-to-Left Posterior-to-Anterior
Inferior-to-Superior (RPI) orientation ::
3dresample -orient RPI
-prefix mprage_RPI.nii.gz
-inset mprage.nii.gz
- Skull-Stripping the image ::
Using AFNI ::
3dSkullStrip -input mprage_RPI.nii.gz
-o_ply mprage_RPI_3dT.nii.gz
or using BET ::
bet mprage_RPI.nii.gz
- The skull-stripping step modifies the intensity values. To get back the
original intensity values, we do an element wise product of RPI data
with step function of skull-stripped data ::
3dcalc -a mprage_RPI.nii.gz
-b mprage_RPI_3dT.nii.gz
-expr 'a*step(b)'
-prefix mprage_RPI_3dc.nii.gz
High Level Workflow Graph:
.. image:: ../images/anatpreproc_graph.dot.png
:width: 500
Detailed Workflow Graph:
.. image:: ../images/anatpreproc_graph_detailed.dot.png
:width: 500
Examples
--------
>>> from CPAC.anat_preproc import create_anat_preproc
>>> preproc = create_anat_preproc()
>>> preproc.inputs.inputspec.anat = 'sub1/anat/mprage.nii.gz'
>>> preproc.run() #doctest: +SKIP
"""
preproc = pe.Workflow(name=wf_name)
inputnode = pe.Node(util.IdentityInterface(fields=['anat',
'brain_mask',
'template_brain_only_for_anat',
'template_skull_for_anat',
'template_brain_only_for_acpc',
'template_skull_for_acpc',
'template_cmass']),
name='inputspec')
outputnode = pe.Node(util.IdentityInterface(fields=['refit',
'reorient',
'skullstrip',
'brain',
'brain_mask',
'anat_skull_leaf',
'center_of_mass']),
name='outputspec')
anat_deoblique = pe.Node(interface=afni.Refit(),
name='anat_deoblique')
anat_deoblique.inputs.deoblique = True
preproc.connect(inputnode, 'anat', anat_deoblique, 'in_file')
preproc.connect(anat_deoblique, 'out_file', outputnode, 'refit')
# Anatomical reorientation
anat_reorient = pe.Node(interface=afni.Resample(),
name='anat_reorient')
anat_reorient.inputs.orientation = 'RPI'
anat_reorient.inputs.outputtype = 'NIFTI_GZ'
preproc.connect(anat_deoblique, 'out_file', anat_reorient, 'in_file')
preproc.connect(anat_reorient, 'out_file', outputnode, 'reorient')
anat_leaf = pe.Node(util.IdentityInterface(fields=['anat_data']),
name='anat_leaf')
if not config.acpc_align:
preproc.connect(anat_reorient, 'out_file', anat_leaf, 'anat_data')
# ACPC alignment
if config.acpc_align:
acpc_align = acpc_alignment(skullstrip_tool=method, config=config, acpc_target=acpc_target, wf_name='acpc_align')
preproc.connect(anat_reorient, 'out_file', acpc_align, 'inputspec.anat_leaf')
preproc.connect(inputnode, 'brain_mask', acpc_align, 'inputspec.brain_mask')
preproc.connect(inputnode, 'template_brain_only_for_acpc', acpc_align, 'inputspec.template_brain_for_acpc')
preproc.connect(inputnode, 'template_skull_for_acpc', acpc_align, 'inputspec.template_head_for_acpc')
preproc.connect(acpc_align, 'outputspec.acpc_aligned_head', anat_leaf, 'anat_data')
if method == 'unet':
preproc.connect(inputnode, 'template_brain_only_for_anat', acpc_align, 'inputspec.template_brain_only_for_anat')
preproc.connect(inputnode, 'template_skull_for_anat', acpc_align, 'inputspec.template_skull_for_anat')
# Disable non_local_means_filtering and n4_bias_field_correction when run niworkflows-ants
if method == 'niworkflows-ants':
config.non_local_means_filtering = False
config.n4_bias_field_correction = False
if config.non_local_means_filtering and config.n4_bias_field_correction:
denoise = pe.Node(interface = ants.DenoiseImage(), name = 'anat_denoise')
preproc.connect(anat_leaf, 'anat_data', denoise, 'input_image')
n4 = pe.Node(interface = ants.N4BiasFieldCorrection(dimension=3, shrink_factor=2, copy_header=True),
name='anat_n4')
preproc.connect(denoise, 'output_image', n4, 'input_image')
elif config.non_local_means_filtering and not config.n4_bias_field_correction:
denoise = pe.Node(interface = ants.DenoiseImage(), name = 'anat_denoise')
preproc.connect(anat_leaf, 'anat_data', denoise, 'input_image')
elif not config.non_local_means_filtering and config.n4_bias_field_correction:
n4 = pe.Node(interface = ants.N4BiasFieldCorrection(dimension=3, shrink_factor=2, copy_header=True),
name='anat_n4')
preproc.connect(anat_leaf, 'anat_data', n4, 'input_image')
anat_leaf2 = pe.Node(util.IdentityInterface(fields=['anat_data']),
name='anat_leaf2')
if config.n4_bias_field_correction:
preproc.connect(n4, 'output_image', anat_leaf2, 'anat_data')
elif config.non_local_means_filtering and not config.n4_bias_field_correction:
preproc.connect(denoise, 'output_image', anat_leaf2, 'anat_data')
else:
preproc.connect(anat_leaf, 'anat_data', anat_leaf2, 'anat_data')
if already_skullstripped:
anat_skullstrip = pe.Node(interface=util.IdentityInterface(fields=['out_file']),
name='anat_skullstrip')
preproc.connect(anat_leaf2, 'anat_data',
anat_skullstrip, 'out_file')
preproc.connect(anat_skullstrip, 'out_file',
outputnode, 'skullstrip')
# binarize skullstripped brain to get brain mask
brain_mask = pe.Node(interface=fsl.maths.MathsCommand(), name='brain_mask')
brain_mask.inputs.args = '-bin'
preproc.connect(anat_skullstrip, 'out_file',
brain_mask, 'in_file')
preproc.connect(brain_mask, 'out_file',
outputnode, 'brain_mask')
preproc.connect(anat_skullstrip, 'out_file',
outputnode, 'brain')
else:
anat_skullstrip = skullstrip_anatomical(method=method, config=config,
wf_name="{0}_skullstrip".format(wf_name))
preproc.connect(anat_leaf2, 'anat_data',
anat_skullstrip, 'inputspec.anat_data')
preproc.connect(inputnode, 'template_brain_only_for_anat',
anat_skullstrip, 'inputspec.template_brain_only_for_anat')
preproc.connect(inputnode, 'template_skull_for_anat',
anat_skullstrip, 'inputspec.template_skull_for_anat')
if method == 'mask' and config.acpc_align:
preproc.connect(acpc_align, 'outputspec.acpc_brain_mask',
anat_skullstrip, 'inputspec.brain_mask')
else:
preproc.connect(inputnode, 'brain_mask',
anat_skullstrip, 'inputspec.brain_mask')
preproc.connect(anat_skullstrip, 'outputspec.brain_mask',
outputnode, 'brain_mask')
preproc.connect(anat_skullstrip, 'outputspec.brain',
outputnode, 'brain')
preproc.connect(anat_leaf2, 'anat_data', outputnode, 'anat_skull_leaf')
return preproc