def create_tensor_pipeline():
"""
Estimate diffusion tensor coefficients and compute FA
:return:
"""
# Nodes
inputnode = pe.Node(
utility.IdentityInterface(fields=["diffusion_volume", "mask"],
mandatory_inputs=False),
name="inputnode",
)
# tensor coefficients estimation
diffusion2tensor = pe.Node(interface=mrtrix3.reconst.FitTensor(),
name="diffusion2tensor")
# derived FA contrast (force nifti format to do registration with FSL or Ants)
tensor2fa = pe.Node(interface=mrtrix3.TensorMetrics(out_fa="fa.nii.gz"),
name="tensor2fa")
outputnode = pe.Node(
utility.IdentityInterface(fields=["tensor_coeff", "fa"],
mandatory_inputs=False),
name="outputnode",
)
# Workflow structure
tensor = pe.Workflow(name="tensor")
tensor.connect(inputnode, "diffusion_volume", diffusion2tensor, "in_file")
tensor.connect(inputnode, "mask", diffusion2tensor, "in_mask")
tensor.connect(diffusion2tensor, "out_file", tensor2fa, "in_file")
tensor.connect(diffusion2tensor, "out_file", outputnode, "tensor_coeff")
tensor.connect(tensor2fa, "out_fa", outputnode, "fa")
return tensor
def calculate_metrics(in_file: Path, out_files: dict):
"""
Calculate various DWI metrics based on Diffusion's kurtosis tensor
estimation.
Parameters
----------
in_file : Path
Path to preprocessed DWI series
out_files : dict
A dictionary containing paths to various metrics to be calculated
"""
tensor = out_files.pop("tensor")
tsr = mrt.FitTensor()
tsr.inputs.in_file = in_file
tsr.inputs.out_file = tensor
tsr.inputs.args = "-quiet"
if not tensor.exists():
tsr.run()
comp = mrt.TensorMetrics()
comp.inputs.in_file = tensor
comp_args = ""
for key, val in out_files.items():
comp_args += f"-{key} {val} "
comp.inputs.args = comp_args
return tsr, comp
def fit_tensors(dwi_file: str, mask_file: str, dti_file: str):
dilated_mask = mask_file.replace(".mif", "_dilated.mif")
cmd = f"maskfilter {mask_file} dilate {dilated_mask} -npass 3"
os.system(cmd)
tsr = mrt.FitTensor()
tsr.inputs.in_file = dwi_file
tsr.inputs.in_mask = dilated_mask
tsr.inputs.out_file = dti_file
print(tsr.cmdline)
tsr.run()
comp = mrt.TensorMetrics()
comp.inputs.in_file = dti_file
comp.inputs.out_fa = dti_file.replace("dti", "fa")
comp.inputs.args = "-force"
print(comp.cmdline)
comp.run()
return comp.inputs.out_fa
def fit_tensors(dwi_file: Path, mask_file: Path, dti_file: Path,
fa_file: Path):
dilated_mask = mask_file.parent / f"{mask_file.stem}_dilated.mif"
cmd = f"maskfilter {mask_file} dilate {dilated_mask} -npass 3"
os.system(cmd)
tsr = mrt.FitTensor()
tsr.inputs.in_file = dwi_file
tsr.inputs.in_mask = dilated_mask
tsr.inputs.out_file = dti_file
print(tsr.cmdline)
tsr.run()
comp = mrt.TensorMetrics()
comp.inputs.in_file = dti_file
comp.inputs.out_fa = fa_file
print(comp.cmdline)
comp.run()
dti_file.unlink()
return comp.inputs.out_fa
# Data conversion from .nii to .mif file (allows to embed diffusion bvals et bvecs)
mrconvert = pe.Node(interface=mrtrix3.MRConvert(), name="mrconvert")
# Bias correction of the diffusion MRI data (for more quantitative approach)
dwibiascorrect = pe.Node(interface=mrtrix3.preprocess.DWIBiasCorrect(),
name="dwibiascorrect")
dwibiascorrect.use_ants = True
# gross brain mask stemming from DWI data
dwi2mask = pe.Node(interface=mrtrix3.utils.BrainMask(), name="dwi2mask")
# tensor coefficients estimation
dwi2tensor = pe.Node(interface=mrtrix3.reconst.FitTensor(), name="dwi2tensor")
# derived FA contrast for registration
tensor2fa = pe.Node(interface=mrtrix3.TensorMetrics(), name="tensor2fa")
# Tissue classification from T1 MRI data
tissue_classif = pe.Node(interface=mrtrix3.Generate5tt(),
name="tissue_classif")
# rely on FSL for T1 tissue segmentation
tissue_classif.inputs.algorithm = "fsl"
# impulsionnal response estimation
dwi2response = pe.Node(interface=mrtrix3.preprocess.ResponseSD(),
name="dwi2response")
dwi2response.inputs.algorithm = "msmt_5tt"
# Multi-shell multi tissue spherical deconvolution of the diffusion MRI data
dwi2fod = pe.Node(
interface=mrtrix3.reconst.ConstrainedSphericalDeconvolution(),
def dholl_preproc_wf(shells=[0, 1000, 2000],
lmax=[0, 8, 8],
sshell=False,
noreorient=False,
template_dir=None,
template_label=None,
wdir=None,
nthreads=1,
name='dholl_preproc_wf'):
"""
Set up Dhollander response preproc workflow
No assumption of registration to T1w space is made
"""
if template_dir is None or template_label is None:
print("Missing template info")
raise IOError
# Grab template data
templateGrabber = io.getTemplate(template_dir=template_dir,
template_label=template_label,
wdir=wdir)
# Convert nii to mif
dwiConvert = pe.Node(mrt.MRConvert(), name='dwiConvert')
dwiConvert.base_dir = wdir
dwiConvert.inputs.nthreads = nthreads
dwiConvert.interface.num_threads = nthreads
# dwi2response - included but not used
dwi2response = pe.Node(mrt.ResponseSD(), name='dwi2response')
dwi2response.base_dir = wdir
dwi2response.inputs.algorithm = 'dhollander'
dwi2response.inputs.wm_file = 'space-dwi_model-CSD_WMResp.txt'
dwi2response.inputs.gm_file = 'space-dwi_model-CSD_GMResp.txt'
dwi2response.inputs.csf_file = 'space-dwi_model-CSD_CSFResp.txt'
dwi2response.inputs.max_sh = lmax
dwi2response.inputs.shell = shells
dwi2response.inputs.nthreads = nthreads
dwi2response.interface.num_threads = nthreads
# Convert mask (nii) to mif
maskConvert = pe.Node(mrt.MRConvert(), name='maskConvert')
maskConvert.base_dir = wdir
maskConvert.inputs.nthreads = nthreads
maskConvert.interface.num_threads = nthreads
# dwi2fod
dwi2fod = pe.Node(mrt.EstimateFOD(), name='dwi2fod')
dwi2fod.base_dir = wdir
dwi2fod.inputs.algorithm = 'msmt_csd'
dwi2fod.inputs.shell = shells
dwi2fod.inputs.wm_odf = 'space-dwi_model-CSD_WMFOD.mif'
if sshell is False:
dwi2fod.inputs.gm_odf = 'space-dwi_model-CSD_GMFOD.mif'
dwi2fod.inputs.csf_odf = 'space-dwi_model-CSD_CSFFOD.mif'
dwi2fod.inputs.nthreads = nthreads
dwi2fod.interface.num_threads = nthreads
# mtnormalise
mtnormalise = pe.Node(mrt.MTNormalise(), name='mtnormalise')
mtnormalise.base_dir = wdir
mtnormalise.inputs.out_wm = 'space-dwi_model-CSD_WMFODNorm.mif'
if sshell is False:
mtnormalise.inputs.out_gm = 'space-dwi_model-CSD_GMFODNorm.mif'
mtnormalise.inputs.out_csf = 'space-dwi_model-CSD_CSFFODNorm.mif'
mtnormalise.inputs.nthreads = nthreads
mtnormalise.interface.num_threads = nthreads
# Registration
MRRegister = pe.Node(mrt.MRRegister(), name='MRRegister')
MRRegister.base_dir = wdir
# MRRegister.inputs.ref_file = template
MRRegister.inputs.nl_warp = [
'from-dwi_to-Template_xfm.mif', 'from-Template_to-dwi_xfm.mif'
]
if noreorient is not False:
MRRegister.inputs.noreorientation = noreorient
MRRegister.inputs.nthreads = nthreads
MRRegister.interface.num_threads = nthreads
# Transforms
WarpSelect1 = pe.Node(niu.Select(), name='WarpSelect1')
WarpSelect1.base_dir = wdir
WarpSelect1.inputs.index = [0]
WarpSelect1.interface.num_threads = nthreads
WarpSelect2 = pe.Node(niu.Select(), name='WarpSelect2')
WarpSelect2.base_dir = wdir
WarpSelect2.inputs.index = [1]
WarpSelect2.interface.num_threads = nthreads
# Warp data
MaskTransform = pe.Node(mrt.MRTransform(), name='MaskTransform')
MaskTransform.base_dir = wdir
MaskTransform.inputs.out_file = 'space-Template_brainmask.mif'
MaskTransform.inputs.nthreads = nthreads
MaskTransform.interface.num_threads = nthreads
FODTransform = pe.Node(mrt.MRTransform(), name='FODTransform')
FODTransform.base_dir = wdir
FODTransform.inputs.out_file = 'space-Template_model-CSD_WMFODNorm.mif'
FODTransform.inputs.nthreads = nthreads
FODTransform.interface.num_threads = nthreads
# Tensor processing
DWINormalise = pe.Node(mrt.DWINormalise(), name='DWINormalise')
DWINormalise.base_dir = wdir
DWINormalise.inputs.out_file = 'space-dwi_dwiNorm.mif'
DWINormalise.inputs.nthreads = nthreads
DWINormalise.interface.num_threads = nthreads
DWITransform = pe.Node(mrt.MRTransform(), name='DWITransform')
DWITransform.base_dir = wdir
DWITransform.inputs.out_file = 'space-Template_dwiNorm.mif'
DWITransform.inputs.nthreads = nthreads
DWITransform.interface.num_threads = nthreads
FitTensor = pe.Node(mrt.FitTensor(), name='FitTensor')
FitTensor.base_dir = wdir
FitTensor.inputs.out_file = 'space-Template_desc-WLS_model-DTI_tensor.mif'
FitTensor.inputs.nthreads = nthreads
FitTensor.interface.num_threads = nthreads
TensorMetrics = pe.Node(mrt.TensorMetrics(), name='TensorMetrics')
TensorMetrics.base_dir = wdir
TensorMetrics.inputs.out_fa = 'space-Template_model-DTI_FA.mif'
TensorMetrics.inputs.out_adc = 'space-Template_model-DTI_MD.mif'
TensorMetrics.inputs.out_ad = 'space-Template_model-DTI_AD.mif'
TensorMetrics.inputs.out_rd = 'space-Template_model-DTI_RD.mif'
TensorMetrics.inputs.nthreads = nthreads
TensorMetrics.interface.num_threads = nthreads
# Build workflow
workflow = pe.Workflow(name=name)
# Single shell
if sshell is True:
workflow.connect([
# Compute FOD
(dwiConvert, dwi2response, [('out_file', 'in_file')]),
(dwiConvert, dwi2fod, [('out_file', 'in_file')]),
(dwi2response, dwi2fod, [('wm_file', 'wm_txt'),
('csf_file', 'csf_txt')]),
(dwi2fod, mtnormalise, [('wm_odf', 'in_wm'),
('csf_odf', 'in_csf')]),
(maskConvert, dwi2response, [('out_file', 'in_mask')]),
(maskConvert, dwi2fod, [('out_file', 'mask_file')]),
(maskConvert, mtnormalise, [('out_file', 'mask')]),
(maskConvert, MRRegister, [('out_file', 'mask1')]),
(templateGrabber, MRRegister, [('wm_fod', 'ref_file'),
('mask', 'mask2')]),
(mtnormalise, MRRegister, [('out_wm', 'in_file')]),
(MRRegister, WarpSelect1, [('nl_warp', 'inlist')]),
(MRRegister, WarpSelect2, [('nl_warp', 'inlist')]),
(maskConvert, MaskTransform, [('out_file', 'in_file')]),
(WarpSelect1, MaskTransform, [('out', 'warp')]),
(mtnormalise, FODTransform, [('out_wm', 'in_file')]),
(WarpSelect1, FODTransform, [('out', 'warp')]),
# Compute tensors
(dwiConvert, DWINormalise, [('out_file', 'in_file')]),
(maskConvert, DWINormalise, [('out_file', 'in_mask')]),
(DWINormalise, DWITransform, [('out_file', 'in_file')]),
(WarpSelect1, DWITransform, [('out', 'warp')]),
(DWITransform, FitTensor, [('out_file', 'in_file')]),
(MaskTransform, FitTensor, [('out_file', 'in_mask')]),
(FitTensor, TensorMetrics, [('out_file', 'in_file')]),
(MaskTransform, TensorMetrics, [('out_file', 'in_mask')])
])
# For multi-shell
else:
workflow.connect([
# Compute FOD
(dwiConvert, dwi2response, [('out_file', 'in_file')]),
(dwiConvert, dwi2fod, [('out_file', 'in_file')]),
(dwi2response, dwi2fod, [('wm_file', 'wm_txt'),
('gm_file', 'gm_txt'),
('csf_file', 'csf_txt')]),
(dwi2fod, mtnormalise, [('wm_odf', 'in_wm'), ('gm_odf', 'in_gm'),
('csf_odf', 'in_csf')]),
(maskConvert, dwi2response, [('out_file', 'in_mask')]),
(maskConvert, dwi2fod, [('out_file', 'mask_file')]),
(maskConvert, mtnormalise, [('out_file', 'mask')]),
(maskConvert, MRRegister, [('out_file', 'mask1')]),
(templateGrabber, MRRegister, [('wm_fod', 'ref_file'),
('mask', 'mask2')]),
(mtnormalise, MRRegister, [('out_wm', 'in_file')]),
(MRRegister, WarpSelect1, [('nl_warp', 'inlist')]),
(MRRegister, WarpSelect2, [('nl_warp', 'inlist')]),
(maskConvert, MaskTransform, [('out_file', 'in_file')]),
(WarpSelect1, MaskTransform, [('out', 'warp')]),
(mtnormalise, FODTransform, [('out_wm', 'in_file')]),
(WarpSelect1, FODTransform, [('out', 'warp')]),
# Compute tensors
(dwiConvert, DWINormalise, [('out_file', 'in_file')]),
(maskConvert, DWINormalise, [('out_file', 'in_mask')]),
(DWINormalise, DWITransform, [('out_file', 'in_file')]),
(WarpSelect1, DWITransform, [('out', 'warp')]),
(DWITransform, FitTensor, [('out_file', 'in_file')]),
(MaskTransform, FitTensor, [('out_file', 'in_mask')]),
(FitTensor, TensorMetrics, [('out_file', 'in_file')]),
(MaskTransform, TensorMetrics, [('out_file', 'in_mask')])
])
return workflow
