def cortex(t1_file, fsdir, outdir, dest_file=None, prefix="cortex",
generate_mask=True, generate_seeds=True):
""" Compute a white matter mask and gyri labelization from the FreeSurfer
'white' surface.
Parameters
----------
t1_file: str (mandatory)
a file containing the t1 image used in FreeSurfer for the segmentation.
fsdir: str( mandatory)
the subject freesurfer segmentation directory.
outdir: str (mandatory)
the output directory.
dest_file: str (optional, default None)
a file containing an image where we want to project the segmentations:
an affine transform is used to align this image to the t1 image.
prefix: str (optional, default 'cortex')
the output files prefix.
generate_mask: bool (optional, default True)
if True generate a white matter binary mask.
generate_seeds: boll (optional, default False)
if True create a 'seeds' directory containing all the gyri mask as
idenpendent files.
Returns
-------
mask_file: str
the white matter mask image file.
label_file: str
the gyri label image file.
seeds: list of str
a list with the seed volumes.
"""
# Create the output directory if necessary
if not os.path.isdir(outdir):
os.makedirs(outdir)
# Load the dataset
t1_image = nibabel.load(t1_file)
t1_affine = t1_image.get_affine()
# If a destination file is specified register it to the t1
if dest_file is not None:
# Load dataset
dest_image = nibabel.load(dest_file)
dest_affine = dest_image.get_affine()
dest_shape = dest_image.get_shape()
# In case of temporal serie extract the first volume
if len(dest_shape) > 3:
temporal_dest_file = dest_file
dest_file = os.path.join(outdir, prefix + "_volume-0.nii.gz")
extract_image(temporal_dest_file, index=0, out_file=dest_file)
dest_shape = dest_shape[:3]
# Register destination image to t1 image
trf_file = os.path.join(outdir, prefix + "_dest_to_t1.trf")
reg_file = os.path.join(outdir, prefix + "_dest_to_t1.nii.gz")
flirt(dest_file, t1_file, omat=trf_file, out=reg_file, usesqform=False,
cost="normmi", dof=6)
voxel_dest_to_t1 = flirt2aff(trf_file, dest_file, t1_file)
voxel_t1_to_dest = numpy.linalg.inv(voxel_dest_to_t1)
# Otherwise use identity transformation
else:
trf_file = None
reg_file = None
dest_affine = t1_affine
dest_shape = t1_image.get_shape()
voxel_t1_to_dest = numpy.identity(4)
# Load the FreeSurfer surface in the 'dest_file' voxel coordinates or
# 't1_file' coordinates if not specified
t1_physical_to_voxel = numpy.linalg.inv(t1_affine)
seg = read_cortex_surface_segmentation(fsdir, t1_physical_to_voxel,
voxel_t1_to_dest)
# Create a mask of the white matter of both hemisphere
if generate_mask:
mask_array = seg["lh"].voxelize(dest_shape)
mask_array += seg["rh"].voxelize(dest_shape)
# Create a gyri label image of both hemisphere
label_array = {}
try:
label_array["lh"], shift_lh = seg["lh"].labelize(dest_shape)
label_array["rh"], shift_rh = seg["rh"].labelize(dest_shape, shift_lh)
except:
if reg_file is not None:
raise FSLResultError("flirt")
raise
# Create the seeds
seeds = []
if generate_seeds:
seedsdir = os.path.join(outdir, "gyri")
if not os.path.isdir(seedsdir):
os.mkdir(seedsdir)
for hemi in ["lh", "rh"]:
surf = seg[hemi]
hemi_label_array = label_array[hemi]
seed_array = numpy.zeros(hemi_label_array.shape,
dtype=hemi_label_array.dtype)
for index, item in surf.metadata.items():
if index != 0:
if hemi == "rh":
index += shift_lh
seed_array[numpy.where(hemi_label_array == index)] = 1
seed_file = os.path.join(
seedsdir,
"{0}-{1}.nii.gz".format(hemi, item["region"]))
seed_image = nibabel.Nifti1Image(seed_array, dest_affine)
nibabel.save(seed_image, seed_file)
seed_array[...] = 0
seeds.append(seed_file)
# Save the mask and label images
mask_file = None
if generate_mask:
mask_file = os.path.join(outdir, prefix + "_mask.nii.gz")
mask_image = nibabel.Nifti1Image(mask_array, dest_affine)
nibabel.save(mask_image, mask_file)
label_array = label_array["lh"] + label_array["rh"]
label_file = os.path.join(outdir, prefix + "_gyri_labels.nii.gz")
label_image = nibabel.Nifti1Image(label_array, dest_affine)
nibabel.save(label_image, label_file)
return mask_file, label_file, seeds, reg_file, trf_file
def anatomical_connectivity_matrix(track_file, label_file, t1_file,
diffusion_file, trf_file, outdir,
symmetric=True):
""" Counts the tracks that start and end at each label pair in the
anatomical space.
Parameters
----------
track_file: str (mandatory)
a text file containing tracks.
label_file: str (mandatory)
a file containing labels that represent a segmentation of the cortex
surface.
t1_file: str (mandatory)
a file containing the t1 image used in FreeSurfer for the segmentation.
diffusion_file: str (optional, default None)
a file containing the diffusion b0 3d image.
trf_file: str (mandatory)
a file with the FSL flirt transformation from the diffusion to the
t1 spaces.
outdir: str (mandatory)
the output directory.
symmetric: bool (optional, default True)
symmetric means we don't distinguish between start and end points. If
symmetric is True, 'matrix[i, j] == matrix[j, i]'.
Returns
-------
matrix: array
the number of connection between each pair of regions defined in the
'label_file'.
"""
# Load the dataset
label_image = nibabel.load(label_file)
label_array = label_image.get_data()
label_shape = label_image.get_shape()
tractogram = Tractogram(track_file)
affine = flirt2aff(trf_file, diffusion_file, t1_file)
# Check the validity of the label array
kind = label_array.dtype.kind
label_positive = (
(kind == "u") or ((kind == "i") and (label_array.min() >= 0)))
if not (label_positive and label_array.ndim == 3):
raise ValueError("Label array must be a 3d integer array with "
"non-negative label values.")
# To compute the connectivity matrix we consider only the first and last
# point of each track
tractogram.apply_affine(affine)
endpoints = tractogram.endpoints().astype(int)
pointsx, pointsy, pointsz = endpoints.T
# Get labels associted to track end points
endlabels = label_array[pointsx, pointsy, pointsz]
if symmetric:
endlabels.sort(axis=0)
matrix = ndbincount(endlabels)
if symmetric:
matrix = numpy.maximum(matrix, matrix.T)
# Remove the connectivity associated to the background
matrix = matrix[1:, 1:]
# Compute the fiber density map
density_map = tractogram.density(shape=label_shape)
# Save the resulting connectivity matrix and density map
proba_file = os.path.join(outdir, "det_paths.nii.gz")
density_image = nibabel.Nifti1Image(density_map, label_image.get_affine())
nibabel.save(density_image, proba_file)
network_file = os.path.join(outdir, "det_network_matrix")
numpy.savetxt(network_file, matrix)
return proba_file, network_file
def cortex(t1_file,
fsdir,
outdir,
dest_file=None,
prefix="cortex",
generate_mask=True,
generate_seeds=True):
""" Compute a white matter mask and gyri labelization from the FreeSurfer
'white' surface.
Parameters
----------
t1_file: str (mandatory)
a file containing the t1 image used in FreeSurfer for the segmentation.
fsdir: str( mandatory)
the subject freesurfer segmentation directory.
outdir: str (mandatory)
the output directory.
dest_file: str (optional, default None)
a file containing an image where we want to project the segmentations:
an affine transform is used to align this image to the t1 image.
prefix: str (optional, default 'cortex')
the output files prefix.
generate_mask: bool (optional, default True)
if True generate a white matter binary mask.
generate_seeds: boll (optional, default False)
if True create a 'seeds' directory containing all the gyri mask as
idenpendent files.
Returns
-------
mask_file: str
the white matter mask image file.
label_file: str
the gyri label image file.
seeds: list of str
a list with the seed volumes.
"""
# Create the output directory if necessary
if not os.path.isdir(outdir):
os.makedirs(outdir)
# Load the dataset
t1_image = nibabel.load(t1_file)
t1_affine = t1_image.get_affine()
# If a destination file is specified register it to the t1
if dest_file is not None:
# Load dataset
dest_image = nibabel.load(dest_file)
dest_affine = dest_image.get_affine()
dest_shape = dest_image.get_shape()
# In case of temporal serie extract the first volume
if len(dest_shape) > 3:
temporal_dest_file = dest_file
dest_file = os.path.join(outdir, prefix + "_volume-0.nii.gz")
extract_image(temporal_dest_file, index=0, out_file=dest_file)
dest_shape = dest_shape[:3]
# Register destination image to t1 image
trf_file = os.path.join(outdir, prefix + "_dest_to_t1.trf")
reg_file = os.path.join(outdir, prefix + "_dest_to_t1.nii.gz")
flirt(dest_file,
t1_file,
omat=trf_file,
out=reg_file,
usesqform=False,
cost="normmi",
dof=6)
voxel_dest_to_t1 = flirt2aff(trf_file, dest_file, t1_file)
voxel_t1_to_dest = numpy.linalg.inv(voxel_dest_to_t1)
# Otherwise use identity transformation
else:
trf_file = None
reg_file = None
dest_affine = t1_affine
dest_shape = t1_image.get_shape()
voxel_t1_to_dest = numpy.identity(4)
# Load the FreeSurfer surface in the 'dest_file' voxel coordinates or
# 't1_file' coordinates if not specified
t1_physical_to_voxel = numpy.linalg.inv(t1_affine)
seg = read_cortex_surface_segmentation(fsdir, t1_physical_to_voxel,
voxel_t1_to_dest)
# Create a mask of the white matter of both hemisphere
if generate_mask:
mask_array = seg["lh"].voxelize(dest_shape)
mask_array += seg["rh"].voxelize(dest_shape)
# Create a gyri label image of both hemisphere
label_array = {}
try:
label_array["lh"], shift_lh = seg["lh"].labelize(dest_shape)
label_array["rh"], shift_rh = seg["rh"].labelize(dest_shape, shift_lh)
except:
if reg_file is not None:
raise FSLResultError("flirt")
raise
# Create the seeds
seeds = []
if generate_seeds:
seedsdir = os.path.join(outdir, "gyri")
if not os.path.isdir(seedsdir):
os.mkdir(seedsdir)
for hemi in ["lh", "rh"]:
surf = seg[hemi]
hemi_label_array = label_array[hemi]
seed_array = numpy.zeros(hemi_label_array.shape,
dtype=hemi_label_array.dtype)
for index, item in surf.metadata.items():
if index != 0:
if hemi == "rh":
index += shift_lh
seed_array[numpy.where(hemi_label_array == index)] = 1
seed_file = os.path.join(
seedsdir,
"{0}-{1}.nii.gz".format(hemi, item["region"]))
seed_image = nibabel.Nifti1Image(seed_array, dest_affine)
nibabel.save(seed_image, seed_file)
seed_array[...] = 0
seeds.append(seed_file)
# Save the mask and label images
mask_file = None
if generate_mask:
mask_file = os.path.join(outdir, prefix + "_mask.nii.gz")
mask_image = nibabel.Nifti1Image(mask_array, dest_affine)
nibabel.save(mask_image, mask_file)
label_array = label_array["lh"] + label_array["rh"]
label_file = os.path.join(outdir, prefix + "_gyri_labels.nii.gz")
label_image = nibabel.Nifti1Image(label_array, dest_affine)
nibabel.save(label_image, label_file)
return mask_file, label_file, seeds, reg_file, trf_file
trf_file = os.path.join(connectdir, "dmri_to_t1.trf")
reg_file = os.path.join(connectdir, "nodif_to_t1.nii.gz")
flirt(nodif_file,
t1_file,
omat=trf_file,
out=reg_file,
usesqform=False,
cost="normmi",
dof=6)
"""
Launch the tractography on the requested point of the cortical surface on the
selected hemisphere
"""
# Load the white mesh in the diffusion space
surface = TriSurface.load(whitefile)
voxel_diff_to_t1 = flirt2aff(trf_file, nodif_file, t1_file)
voxel_t1_to_diff = numpy.linalg.inv(voxel_diff_to_t1)
white_diff_vertices = apply_affine_on_mesh(surface.vertices, voxel_t1_to_diff)
# Select the vertices of interest
if vertices_indices is None:
vertices_indices = range(len(surface.vertices))
# Go through all the hemisphere vertices
textures = {}
for index in vertices_indices:
# Select the seeding vertex
point = white_diff_vertices[index]
# Create a directory for each seeding vertex in order to avoid collision
to get it.
"""
if trf_file is None:
trf_file = os.path.join(connectdir, "dmri_to_t1.trf")
reg_file = os.path.join(connectdir, "nodif_to_t1.nii.gz")
flirt(nodif_file, t1_file, omat=trf_file, out=reg_file, usesqform=False,
cost="normmi", dof=6)
"""
Launch the tractography on the requested point of the cortical surface on the
selected hemisphere
"""
# Load the white mesh in the diffusion space
surface = TriSurface.load(whitefile)
voxel_diff_to_t1 = flirt2aff(trf_file, nodif_file, t1_file)
voxel_t1_to_diff = numpy.linalg.inv(voxel_diff_to_t1)
white_diff_vertices = apply_affine_on_mesh(surface.vertices, voxel_t1_to_diff)
# Select the vertices of interest
if vertices_indices is None:
vertices_indices = range(len(surface.vertices))
# Go through all the hemisphere vertices
textures = {}
for index in vertices_indices:
# Select the seeding vertex
point = white_diff_vertices[index]
# Create a directory for each seeding vertex in order to avoid collision
def anatomical_connectivity_matrix(track_file,
label_file,
t1_file,
diffusion_file,
trf_file,
outdir,
symmetric=True):
""" Counts the tracks that start and end at each label pair in the
anatomical space.
Parameters
----------
track_file: str (mandatory)
a text file containing tracks.
label_file: str (mandatory)
a file containing labels that represent a segmentation of the cortex
surface.
t1_file: str (mandatory)
a file containing the t1 image used in FreeSurfer for the segmentation.
diffusion_file: str (optional, default None)
a file containing the diffusion b0 3d image.
trf_file: str (mandatory)
a file with the FSL flirt transformation from the diffusion to the
t1 spaces.
outdir: str (mandatory)
the output directory.
symmetric: bool (optional, default True)
symmetric means we don't distinguish between start and end points. If
symmetric is True, 'matrix[i, j] == matrix[j, i]'.
Returns
-------
matrix: array
the number of connection between each pair of regions defined in the
'label_file'.
"""
# Load the dataset
label_image = nibabel.load(label_file)
label_array = label_image.get_data()
label_shape = label_image.get_shape()
tractogram = Tractogram(track_file)
affine = flirt2aff(trf_file, diffusion_file, t1_file)
# Check the validity of the label array
kind = label_array.dtype.kind
label_positive = ((kind == "u")
or ((kind == "i") and (label_array.min() >= 0)))
if not (label_positive and label_array.ndim == 3):
raise ValueError("Label array must be a 3d integer array with "
"non-negative label values.")
# To compute the connectivity matrix we consider only the first and last
# point of each track
tractogram.apply_affine(affine)
endpoints = tractogram.endpoints().astype(int)
pointsx, pointsy, pointsz = endpoints.T
# Get labels associted to track end points
endlabels = label_array[pointsx, pointsy, pointsz]
if symmetric:
endlabels.sort(axis=0)
matrix = ndbincount(endlabels)
if symmetric:
matrix = numpy.maximum(matrix, matrix.T)
# Remove the connectivity associated to the background
matrix = matrix[1:, 1:]
# Compute the fiber density map
density_map = tractogram.density(shape=label_shape)
# Save the resulting connectivity matrix and density map
proba_file = os.path.join(outdir, "det_paths.nii.gz")
density_image = nibabel.Nifti1Image(density_map, label_image.get_affine())
nibabel.save(density_image, proba_file)
network_file = os.path.join(outdir, "det_network_matrix")
numpy.savetxt(network_file, matrix)
return proba_file, network_file
