def ThresholdImage(volume, output_file='', threshlo=1, threshhi=10000):
"""
Use the ThresholdImage function in ANTs to threshold image volume.
Usage: ThresholdImage ImageDimension ImageIn.ext outImage.ext
threshlo threshhi <insideValue> <outsideValue>
Parameters
----------
volume : string
nibabel-readable image volume
output_file : string
nibabel-readable image volume
threshlo : integer
lower threshold
threshhi : integer
upper threshold
Returns
-------
output_file : string
name of output nibabel-readable image volume
Examples
--------
>>> import os
>>> from mindboggle.thirdparty.ants import ThresholdImage
>>> from mindboggle.mio.fetch_data import prep_tests
>>> urls, fetch_data = prep_tests()
>>> volume = fetch_data(urls['T1_001'], '', '.nii.gz')
>>> os.rename(volume, volume + '.nii.gz')
>>> volume += '.nii.gz'
>>> output_file = ''
>>> threshlo = 500
>>> threshhi = 10000
>>> output_file = ThresholdImage(volume, output_file, threshlo, threshhi) # doctest: +SKIP
View result (skip test):
>>> from mindboggle.mio.plots import plot_volumes
>>> plot_volumes(output_file) # doctest: +SKIP
"""
import os
from mindboggle.guts.utilities import execute
if not output_file:
output_file = os.path.join(os.getcwd(),
'threshold_' + os.path.basename(volume))
cmd = [
'ThresholdImage', '3', volume, output_file,
str(threshlo),
str(threshhi)
]
execute(cmd, 'os')
if not os.path.exists(output_file):
raise IOError("ThresholdImage did not create " + output_file + ".")
return output_file
def ThresholdImage(volume, output_file='', threshlo=1, threshhi=10000):
"""
Use the ThresholdImage function in ANTs to threshold image volume.
Usage: ThresholdImage ImageDimension ImageIn.ext outImage.ext
threshlo threshhi <insideValue> <outsideValue>
Parameters
----------
volume : string
nibabel-readable image volume
output_file : string
nibabel-readable image volume
threshlo : integer
lower threshold
threshhi : integer
upper threshold
Returns
-------
output_file : string
name of output nibabel-readable image volume
Examples
--------
>>> import os
>>> from mindboggle.thirdparty.ants import ThresholdImage
>>> from mindboggle.mio.fetch_data import prep_tests
>>> urls, fetch_data = prep_tests()
>>> volume = fetch_data(urls['T1_001'])
>>> os.rename(volume, volume + '.nii.gz')
>>> volume += '.nii.gz'
>>> output_file = ''
>>> threshlo = 500
>>> threshhi = 10000
>>> output_file = ThresholdImage(volume, output_file, threshlo, threshhi) # doctest: +SKIP
View result (skip test):
>>> from mindboggle.mio.plots import plot_volumes
>>> plot_volumes(output_file) # doctest: +SKIP
"""
import os
from mindboggle.guts.utilities import execute
if not output_file:
output_file = os.path.join(os.getcwd(),
'threshold_' + os.path.basename(volume))
cmd = ['ThresholdImage', '3', volume, output_file,
str(threshlo), str(threshhi)]
execute(cmd, 'os')
if not os.path.exists(output_file):
raise IOError("ThresholdImage did not create " + output_file + ".")
return output_file
def ThresholdImage(volume, output_file='', threshlo=1, threshhi=10000):
"""
Use the ThresholdImage function in ANTs to threshold image volume::
Usage: ThresholdImage ImageDimension ImageIn.ext outImage.ext
threshlo threshhi <insideValue> <outsideValue>
Parameters
----------
volume : string
nibabel-readable image volume
output_file : string
nibabel-readable image volume
threshlo : integer
lower threshold
threshhi : integer
upper threshold
Returns
-------
output_file : string
name of output nibabel-readable image volume
Examples
--------
>>> import os
>>> from mindboggle.thirdparty.ants import ThresholdImage
>>> from mindboggle.mio.plots import plot_volumes
>>> path = os.path.join(os.environ['MINDBOGGLE_DATA'])
>>> volume = os.path.join(path, 'arno', 'mri', 't1weighted.nii.gz')
>>> output_file = ''
>>> threshlo = 500
>>> threshhi = 10000
>>> output_file = ThresholdImage(volume, output_file, threshlo, threshhi)
>>> # View
>>> plot_volumes(output_file)
"""
import os
from mindboggle.guts.utilities import execute
if not output_file:
output_file = os.path.join(os.getcwd(),
'threshold_' + os.path.basename(volume))
cmd = 'ThresholdImage 3 {0} {1} {2} {3}'.format(volume, output_file,
threshlo, threshhi)
execute(cmd, 'os')
if not os.path.exists(output_file):
raise (IOError("ThresholdImage did not create " + output_file + "."))
return output_file
def ThresholdImage(volume, output_file='', threshlo=1, threshhi=10000):
"""
Use the ThresholdImage function in ANTs to threshold image volume::
Usage: ThresholdImage ImageDimension ImageIn.ext outImage.ext
threshlo threshhi <insideValue> <outsideValue>
Parameters
----------
volume : string
nibabel-readable image volume
output_file : string
nibabel-readable image volume
threshlo : integer
lower threshold
threshhi : integer
upper threshold
Returns
-------
output_file : string
name of output nibabel-readable image volume
Examples
--------
>>> import os
>>> from mindboggle.thirdparty.ants import ThresholdImage
>>> from mindboggle.mio.plots import plot_volumes
>>> path = os.path.join(os.environ['MINDBOGGLE_DATA'])
>>> volume = os.path.join(path, 'arno', 'mri', 't1weighted.nii.gz')
>>> output_file = ''
>>> threshlo = 500
>>> threshhi = 10000
>>> output_file = ThresholdImage(volume, output_file, threshlo, threshhi)
>>> # View
>>> plot_volumes(output_file)
"""
import os
from mindboggle.guts.utilities import execute
if not output_file:
output_file = os.path.join(os.getcwd(),
'threshold_' + os.path.basename(volume))
cmd = 'ThresholdImage 3 {0} {1} {2} {3}'.format(volume, output_file,
threshlo, threshhi)
execute(cmd, 'os')
if not os.path.exists(output_file):
raise(IOError("ThresholdImage did not create " + output_file + "."))
return output_file
def plot_volumes(volume_files, command='fslview'):
"""
Use fslview to visualize image volume data.
Parameters
----------
volume_files : list of strings
names of image volume files
command : string
plotting software command
Examples
--------
>>> import os
>>> from mindboggle.mio.plots import plot_volumes
>>> from mindboggle.mio.fetch_data import prep_tests
>>> urls, fetch_data = prep_tests()
>>> label_file1 = fetch_data(urls['freesurfer_labels'], '', '.vtk')
>>> label_file2 = fetch_data(urls['freesurfer_labels'], '', '.vtk')
>>> volume_files = [label_file1, label_file2]
>>> command = 'fslview'
>>> command = '/Applications/ITK-SNAP.app/Contents/MacOS/InsightSNAP'
>>> plot_volumes(volume_files, command=command) # doctest: +SKIP
"""
from mindboggle.guts.utilities import execute
if isinstance(volume_files, str):
volume_files = [volume_files]
elif not isinstance(volume_files, list):
raise IOError('plot_volumes() requires volume_files to be a list '
'or string.')
if not isinstance(command, str):
raise IOError('plot_volumes() requires command to be a string.')
else:
command = [command]
command.extend(volume_files)
command.extend('&')
execute(command, 'os')
def plot_volumes(volume_files, command='fslview'):
"""
Use fslview to visualize image volume data.
Parameters
----------
volume_files : list of strings
names of image volume files
command : string
plotting software command
Examples
--------
>>> import os
>>> from mindboggle.mio.plots import plot_volumes
>>> from mindboggle.mio.fetch_data import prep_tests
>>> urls, fetch_data = prep_tests()
>>> label_file1 = fetch_data(urls['freesurfer_labels'], '', '.vtk')
>>> label_file2 = fetch_data(urls['freesurfer_labels'], '', '.vtk')
>>> volume_files = [label_file1, label_file2]
>>> command = 'fslview'
>>> command = '/Applications/ITK-SNAP.app/Contents/MacOS/InsightSNAP'
>>> plot_volumes(volume_files, command=command) # doctest: +SKIP
"""
from mindboggle.guts.utilities import execute
if isinstance(volume_files, str):
volume_files = [volume_files]
elif not isinstance(volume_files, list):
raise IOError('plot_volumes() requires volume_files to be a list '
'or string.')
if not isinstance(command, str):
raise IOError('plot_volumes() requires command to be a string.')
else:
command = [command]
command.extend(volume_files)
command.extend('&')
execute(command, 'os')
def plot_volumes(volume_files, command='fslview'):
"""
Use fslview to visualize image volume data.
Parameters
----------
volume_files : list of strings
names of image volume files
command : string
plotting software command
Examples
--------
>>> import os
>>> from mindboggle.mio.plots import plot_volumes
>>> path = os.environ['MINDBOGGLE_DATA']
>>> volume_file1 = os.path.join(path, 'Twins-2-1', 'mri', 't1weighted.nii.gz')
>>> volume_file2 = os.path.join(path, 'Twins-2-1', 'mri', 't1weighted_brain.nii.gz')
>>> volume_files = [volume_file1, volume_file2]
>>> command = 'fslview'
>>> command = '/Applications/ITK-SNAP.app/Contents/MacOS/InsightSNAP'
>>> plot_volumes(volume_files, command=command)
"""
from mindboggle.guts.utilities import execute
if isinstance(volume_files, str):
volume_files = [volume_files]
elif not isinstance(volume_files, list):
raise(IOError('plot_volumes() requires volume_files to be a list or string.'))
if not isinstance(command, str):
raise(IOError('plot_volumes() requires command to be a string.'))
else:
command = [command]
command.extend(volume_files)
command.extend('&')
execute(command, 'os')
def ResampleImageBySpacing(volume,
output_file='',
outxspc=1,
outyspc=1,
outzspc=1,
dosmooth=0,
addvox=0,
nninterp=1):
"""
Use the ResampleImageBySpacing function in ANTs to resample image volume.
Usage: ResampleImageBySpacing ImageDimension ImageIn.ext outImage.ex
outxspc outyspc <outzspc> <dosmooth?> <addvox> <nninterp?>
Parameters
----------
volume : string
nibabel-readable image volume
output_file : string
nibabel-readable image volume
outxspc : integer
output x-spacing
outyspc : integer
output y-spacing
outzspc : integer
output z-spacing
dosmooth : bool
smooth?
addvox : integer
pad each dimension by addvox
nninterp : bool
nearest-neighbor interpolation?
Returns
-------
output_file : string
name of output nibabel-readable image volume
Examples
--------
>>> # Resample image so that 1mm voxels are 0.5mm voxels:
>>> import os
>>> from mindboggle.thirdparty.ants import ResampleImageBySpacing
>>> from mindboggle.mio.fetch_data import prep_tests
>>> urls, fetch_data = prep_tests()
>>> volume = fetch_data(urls['T1_001'], '', '.nii.gz')
>>> os.rename(volume, volume + '.nii.gz')
>>> volume += '.nii.gz'
>>> output_file = ''
>>> outxspc = 1/2.0
>>> outyspc = 1/2.0
>>> outzspc = 1/2.0
>>> dosmooth = 0
>>> addvox = 0
>>> nninterp = 1
>>> output_file = ResampleImageBySpacing(volume, output_file, outxspc,
... outxspc, outzspc, dosmooth, addvox, nninterp) # doctest: +SKIP
View result (skip test):
>>> from mindboggle.mio.plots import plot_volumes
>>> plot_volumes(output_file) # doctest: +SKIP
"""
import os
from mindboggle.guts.utilities import execute
if not output_file:
output_file = os.path.join(os.getcwd(),
'resampled_' + os.path.basename(volume))
cmd = [
'ResampleImageBySpacing', '3', volume, output_file,
str(outxspc),
str(outyspc),
str(outzspc)
]
execute(cmd, 'os')
if not os.path.exists(output_file):
raise IOError(
"ResampleImageBySpacing did not create {0).".format(output_file))
return output_file
def PropagateLabelsThroughMask(mask,
labels,
mask_index=None,
output_file='',
binarize=True,
stopvalue=''):
"""
Use ANTs to fill a binary volume mask with initial labels.
This program uses ThresholdImage and the ImageMath
PropagateLabelsThroughMask functions in ANTs.
ThresholdImage ImageDimension ImageIn.ext outImage.ext
threshlo threshhi <insideValue> <outsideValue>
PropagateLabelsThroughMask: Final output is the propagated label image.
ImageMath ImageDimension Out.ext PropagateLabelsThroughMask
speed/binaryimagemask.nii.gz initiallabelimage.nii.gz ...
Parameters
----------
mask : string
nibabel-readable image volume
labels : string
nibabel-readable image volume with integer labels
mask_index : integer (optional)
mask with just voxels having this value
output_file : string (optional)
nibabel-readable labeled image volume
binarize : bool (optional)
binarize mask?
stopvalue : integer (optional)
stopping value
Returns
-------
output_file : string
name of labeled output nibabel-readable image volume
Examples
--------
>>> # Propagate FreeSurfer labels through brain mask:
>>> import os
>>> from mindboggle.thirdparty.ants import PropagateLabelsThroughMask
>>> from mindboggle.mio.fetch_data import prep_tests
>>> urls, fetch_data = prep_tests()
>>> labels = fetch_data(urls['freesurfer_labels'], '', '.nii.gz')
>>> mask = fetch_data(urls['ants_mask'], '', '.nii.gz')
>>> mask_index = None
>>> output_file = ''
>>> binarize = True
>>> stopvalue = None
>>> output_file = PropagateLabelsThroughMask(mask, labels, mask_index,
... output_file, binarize, stopvalue) # doctest: +SKIP
View result (skip test):
>>> from mindboggle.mio.plots import plot_volumes
>>> plot_volumes(output_file) # doctest: +SKIP
"""
import os
from mindboggle.guts.utilities import execute
if not output_file:
#output_file = os.path.join(os.getcwd(),
# 'PropagateLabelsThroughMask.nii.gz')
output_file = os.path.join(
os.getcwd(),
os.path.basename(labels) + '_through_' + os.path.basename(mask))
print('mask: {0}, labels: {1}'.format(mask, labels))
# Binarize image volume:
if binarize:
temp_file = os.path.join(os.getcwd(),
'PropagateLabelsThroughMask.nii.gz')
cmd = ['ThresholdImage', '3', mask, temp_file, '0 1 0 1']
execute(cmd, 'os')
mask = temp_file
# Mask with just voxels having mask_index value:
if mask_index:
mask2 = os.path.join(os.getcwd(), 'temp.nii.gz')
cmd = [
'ThresholdImage', '3', mask, mask2,
str(mask_index),
str(mask_index)
]
execute(cmd, 'os')
else:
mask2 = mask
# Propagate labels:
cmd = [
'ImageMath', '3', output_file, 'PropagateLabelsThroughMask', mask2,
labels
]
if stopvalue:
cmd.extend(str(stopvalue))
execute(cmd, 'os')
if not os.path.exists(output_file):
raise IOError("ImageMath did not create " + output_file + ".")
return output_file
def antsApplyTransformsToPoints(points, transform_files, inverse_booleans=[0]):
"""
Run ANTs antsApplyTransformsToPoints function to transform points.
(Creates pre- and post-transformed .csv points files for ANTs.)
Parameters
----------
points : list of lists of three integers
point coordinate data
transform_files : list
transform file names
inverse_booleans : list
for each transform, one to apply inverse of transform (otherwise zero)
Returns
-------
transformed_points : list of lists of three integers
transformed point coordinate data
Examples
--------
>>> import numpy as np
>>> from mindboggle.thirdparty.ants import antsApplyTransformsToPoints
>>> from mindboggle.mio.vtks import read_points
>>> from mindboggle.mio.fetch_data import prep_tests
>>> urls, fetch_data = prep_tests()
>>> xfm1 = fetch_data(urls['ants_affine_template2subject'], '', '.mat')
>>> xfm2 = fetch_data(urls['ants_warp_template2subject'], '', '.nii.gz')
>>> xfm3 = fetch_data(urls['OASIS-30_Atropos_template_to_MNI152_affine'], '', '.txt')
>>> transform_files = [xfm1, xfm2, xfm3]
>>> vtk_file = fetch_data(urls['left_pial'], '', '.vtk')
>>> points = read_points(vtk_file)
>>> inverse_booleans = [0,0,1]
>>> transformed_points = antsApplyTransformsToPoints(points,
... transform_files, inverse_booleans) # doctest: +SKIP
>>> print(np.array_str(np.array(transformed_points[0:5]),
... precision=5, suppress_small=True)) # doctest: +SKIP
[[-11.23189 -46.78223 -39.88869]
[-11.71384 -46.87075 -40.13328]
[-12.56237 -46.99126 -40.04564]
[ -9.66693 -46.0446 -41.36334]
[-10.67998 -46.45458 -40.7572 ]]
"""
import os
from io import open
from mindboggle.guts.utilities import execute
# ------------------------------------------------------------------------
# Write points (x,y,z,1) to a .csv file:
# ------------------------------------------------------------------------
points_file = os.path.join(os.getcwd(), 'points.csv')
fid = open(points_file, 'w')
fid.write('x,y,z,t\n')
fid.close()
fid = open(points_file, 'a')
for point in points:
string_of_zeros = (4 - len(point)) * ',0'
fid.write(','.join([str(x) for x in point]) + string_of_zeros + '\n')
fid.close()
# ------------------------------------------------------------------------
# Apply transforms to points in .csv file:
# ------------------------------------------------------------------------
transformed_points_file = os.path.join(os.getcwd(),
'transformed_points.csv')
transform_string = ''
for ixfm, transform_file in enumerate(transform_files):
transform_string += " --t [{0},{1}]".\
format(transform_file, str(inverse_booleans[ixfm]))
cmd = [
'antsApplyTransformsToPoints', '-d', '3', '-i', points_file, '-o',
transformed_points_file, transform_string
]
try:
execute(cmd, 'os')
except:
raise Exception("Cannot find antsApplyTransformsToPoints command.")
if not os.path.exists(transformed_points_file):
raise IOError("antsApplyTransformsToPoints did not create {0}.".format(
transformed_points_file))
# ------------------------------------------------------------------------
# Return transformed points:
# ------------------------------------------------------------------------
fid = open(transformed_points_file, 'r')
lines = fid.readlines()
fid.close()
transformed_points = []
for iline, line in enumerate(lines):
if iline > 0:
point_xyz1 = [float(x) for x in line.split(',')]
transformed_points.append(point_xyz1[0:3])
return transformed_points
def PropagateLabelsThroughMask(mask,
labels,
mask_index=None,
output_file='',
binarize=True,
stopvalue=''):
"""
Use ANTs to fill a binary volume mask with initial labels.
This program uses ThresholdImage and the ImageMath
PropagateLabelsThroughMask functions in ANTs.
ThresholdImage ImageDimension ImageIn.ext outImage.ext
threshlo threshhi <insideValue> <outsideValue>
PropagateLabelsThroughMask: Final output is the propagated label image.
ImageMath ImageDimension Out.ext PropagateLabelsThroughMask
speed/binaryimagemask.nii.gz initiallabelimage.nii.gz ...
Parameters
----------
mask : string
nibabel-readable image volume
labels : string
nibabel-readable image volume with integer labels
mask_index : integer (optional)
mask with just voxels having this value
output_file : string
nibabel-readable labeled image volume
binarize : Boolean
binarize mask?
stopvalue : integer
stopping value
Returns
-------
output_file : string
name of labeled output nibabel-readable image volume
Examples
--------
>>> import os
>>> from mindboggle.thirdparty.ants import PropagateLabelsThroughMask
>>> from mindboggle.mio.plots import plot_volumes
>>> path = os.path.join(os.environ['MINDBOGGLE_DATA'])
>>> labels = os.path.join(path, 'arno', 'labels', 'labels.DKT25.manual.nii.gz')
>>> mask = os.path.join(path, 'arno', 'mri', 't1weighted_brain.nii.gz')
>>> mask_index = None
>>> output_file = ''
>>> binarize = True
>>> stopvalue = None
>>> output_file = PropagateLabelsThroughMask(mask, labels, mask_index, output_file, binarize, stopvalue)
>>> # View
>>> plot_volumes(output_file)
"""
import os
from mindboggle.guts.utilities import execute
if not output_file:
#output_file = os.path.join(os.getcwd(),
# 'PropagateLabelsThroughMask.nii.gz')
output_file = os.path.join(
os.getcwd(),
os.path.basename(labels) + '_through_' + os.path.basename(mask))
print('mask: {0}, labels: {1}'.format(mask, labels))
# Binarize image volume:
if binarize:
temp_file = os.path.join(os.getcwd(),
'PropagateLabelsThroughMask.nii.gz')
cmd = ['ThresholdImage', '3', mask, temp_file, '0 1 0 1']
execute(cmd, 'os')
mask = temp_file
# Mask with just voxels having mask_index value:
if mask_index:
mask2 = os.path.join(os.getcwd(), 'temp.nii.gz')
cmd = 'ThresholdImage 3 {0} {1} {2} {3} 1 0'.format(
mask, mask2, mask_index, mask_index)
execute(cmd)
else:
mask2 = mask
# Propagate labels:
cmd = [
'ImageMath', '3', output_file, 'PropagateLabelsThroughMask', mask2,
labels
]
if stopvalue:
cmd.extend(stopvalue)
execute(cmd, 'os')
if not os.path.exists(output_file):
raise (IOError("ImageMath did not create " + output_file + "."))
return output_file
def ImageMath(volume1, volume2, operator='m', output_file=''):
"""
Use the ImageMath function in ANTs to perform operation on two volumes::
m : Multiply --- use vm for vector multiply
+ : Add --- use v+ for vector add
- : Subtract --- use v- for vector subtract
/ : Divide
^ : Power
exp : Take exponent exp(imagevalue*value)
addtozero : add image-b to image-a only over points where image-a has zero values
overadd : replace image-a pixel with image-b pixel if image-b pixel is non-zero
abs : absolute value
total : Sums up values in an image or in image1*image2 (img2 is the probability mask)
mean : Average of values in an image or in image1*image2 (img2 is the probability mask)
vtotal : Sums up volumetrically weighted values in an image or in image1*image2 (img2 is the probability mask)
Decision : Computes result=1./(1.+exp(-1.0*( pix1-0.25)/pix2))
Neg : Produce image negative
Parameters
----------
volume1 : string
nibabel-readable image volume
volume2 : string
nibabel-readable image volume
operator : string
ImageMath string corresponding to mathematical operator
output_file : string
nibabel-readable image volume
Returns
-------
output_file : string
name of output nibabel-readable image volume
Examples
--------
>>> # Mask head with brain mask:
>>> import os
>>> from mindboggle.thirdparty.ants import ImageMath
>>> from mindboggle.mio.fetch_data import prep_tests
>>> urls, fetch_data = prep_tests()
>>> volume1 = fetch_data(urls['T1_001'], '', '.nii.gz')
>>> volume2 = fetch_data(urls['ants_mask'], '', '.nii.gz')
>>> operator = 'm'
>>> output_file = ''
>>> output_file = ImageMath(volume1, volume2, operator, output_file) # doctest: +SKIP
View result (skip test):
>>> from mindboggle.mio.plots import plot_volumes
>>> plot_volumes(output_file) # doctest: +SKIP
"""
import os
from mindboggle.guts.utilities import execute
if not output_file:
output_file = os.path.join(
os.getcwd(),
os.path.basename(volume1) + '_' + os.path.basename(volume2))
cmd = ['ImageMath', '3', output_file, operator, volume1, volume2]
execute(cmd, 'os')
if not os.path.exists(output_file):
raise IOError("ImageMath did not create " + output_file + ".")
return output_file
def ResampleImageBySpacing(volume, output_file='', outxspc=1, outyspc=1,
outzspc=1, dosmooth=0, addvox=0, nninterp=1):
"""
Use the ResampleImageBySpacing function in ANTs to resample image volume.
Usage: ResampleImageBySpacing ImageDimension ImageIn.ext outImage.ex
outxspc outyspc <outzspc> <dosmooth?> <addvox> <nninterp?>
Parameters
----------
volume : string
nibabel-readable image volume
output_file : string
nibabel-readable image volume
outxspc : integer
output x-spacing
outyspc : integer
output y-spacing
outzspc : integer
output z-spacing
dosmooth : bool
smooth?
addvox : integer
pad each dimension by addvox
nninterp : bool
nearest-neighbor interpolation?
Returns
-------
output_file : string
name of output nibabel-readable image volume
Examples
--------
>>> # Resample image so that 1mm voxels are 0.5mm voxels:
>>> import os
>>> from mindboggle.thirdparty.ants import ResampleImageBySpacing
>>> from mindboggle.mio.fetch_data import prep_tests
>>> urls, fetch_data = prep_tests()
>>> volume = fetch_data(urls['T1_001'])
>>> os.rename(volume, volume + '.nii.gz')
>>> volume += '.nii.gz'
>>> output_file = ''
>>> outxspc = 1/2.0
>>> outyspc = 1/2.0
>>> outzspc = 1/2.0
>>> dosmooth = 0
>>> addvox = 0
>>> nninterp = 1
>>> output_file = ResampleImageBySpacing(volume, output_file, outxspc,
... outxspc, outzspc, dosmooth, addvox, nninterp) # doctest: +SKIP
View result (skip test):
>>> from mindboggle.mio.plots import plot_volumes
>>> plot_volumes(output_file) # doctest: +SKIP
"""
import os
from mindboggle.guts.utilities import execute
if not output_file:
output_file = os.path.join(os.getcwd(),
'resampled_' + os.path.basename(volume))
cmd = ['ResampleImageBySpacing', '3', volume, output_file,
str(outxspc), str(outyspc), str(outzspc)]
execute(cmd, 'os')
if not os.path.exists(output_file):
raise IOError("ResampleImageBySpacing did not create {0).".
format(output_file))
return output_file
def plot_mask_surface(vtk_file, mask_file='', nonmask_value=-1,
masked_output='', remove_nonmask=False,
program='vtkviewer',
use_colormap=False, colormap_file=''):
"""
Use vtkviewer or mayavi2 to visualize VTK surface mesh data.
If a mask_file is provided, a temporary masked file is saved,
and it is this file that is viewed.
If using vtkviewer, can optionally provide colormap file
or set $COLORMAP environment variable.
Parameters
----------
vtk_file : string
name of VTK surface mesh file
mask_file : string
name of VTK surface mesh file to mask vtk_file vertices
nonmask_value : integer
nonmask (usually background) value
masked_output : string
temporary masked output file name
remove_nonmask : Boolean
remove vertices that are not in mask? (otherwise assign nonmask_value)
program : string {'vtkviewer', 'mayavi2'}
program to visualize VTK file
use_colormap : Boolean
use Paraview-style XML colormap file set by $COLORMAP env variable?
colormap_file : string
use colormap in given file if use_colormap==True? if empty and
use_colormap==True, use file set by $COLORMAP environment variable
Examples
--------
>>> import os
>>> from mindboggle.mio.plots import plot_mask_surface
>>> path = os.environ['MINDBOGGLE_DATA']
>>> vtk_file = os.path.join(path, 'arno', 'labels', 'lh.labels.DKT31.manual.vtk')
>>> mask_file = os.path.join(path, 'test_one_label.vtk')
>>> nonmask_value = 0 #-1
>>> masked_output = ''
>>> remove_nonmask = True
>>> program = 'vtkviewer'
>>> use_colormap = True
>>> colormap_file = '' #'/software/surface_cpp_tools/colormap.xml'
>>> plot_mask_surface(vtk_file, mask_file, nonmask_value, masked_output, remove_nonmask, program, use_colormap, colormap_file)
"""
import os
import numpy as np
from mindboggle.guts.mesh import remove_faces, reindex_faces_points
from mindboggle.guts.utilities import execute
from mindboggle.mio.plots import plot_surfaces
from mindboggle.mio.vtks import read_scalars, rewrite_scalars, \
read_vtk, write_vtk
#-------------------------------------------------------------------------
# Filter mesh with non-background values from a second (same-size) mesh:
#-------------------------------------------------------------------------
if mask_file:
mask, name = read_scalars(mask_file, True, True)
if not masked_output:
masked_output = os.path.join(os.getcwd(), 'temp.vtk')
file_to_plot = masked_output
#---------------------------------------------------------------------
# Remove nonmask-valued vertices:
#---------------------------------------------------------------------
if remove_nonmask:
#-----------------------------------------------------------------
# Load VTK files:
#-----------------------------------------------------------------
points, indices, lines, faces, scalars, scalar_names, npoints, \
input_vtk = read_vtk(vtk_file, True, True)
#-----------------------------------------------------------------
# Find mask indices, remove nonmask faces, and reindex:
#-----------------------------------------------------------------
Imask = [i for i,x in enumerate(mask) if x != nonmask_value]
mask_faces = remove_faces(faces, Imask)
mask_faces, points, \
original_indices = reindex_faces_points(mask_faces, points)
#-----------------------------------------------------------------
# Write VTK file with scalar values:
#-----------------------------------------------------------------
if np.ndim(scalars) == 1:
scalar_type = type(scalars[0]).__name__
elif np.ndim(scalars) == 2:
scalar_type = type(scalars[0][0]).__name__
else:
print("Undefined scalar type!")
write_vtk(file_to_plot, points, [], [], mask_faces,
scalars[original_indices].tolist(), scalar_names,
scalar_type=scalar_type)
else:
scalars, name = read_scalars(vtk_file, True, True)
scalars[mask == nonmask_value] = nonmask_value
rewrite_scalars(vtk_file, file_to_plot, scalars)
else:
file_to_plot = vtk_file
#-------------------------------------------------------------------------
# Display with vtkviewer.py:
#-------------------------------------------------------------------------
if program == 'vtkviewer':
plot_surfaces(file_to_plot, use_colormap=use_colormap,
colormap_file=colormap_file)
#-------------------------------------------------------------------------
# Display with mayavi2:
#-------------------------------------------------------------------------
elif program == 'mayavi2':
cmd = ["mayavi2", "-d", file_to_plot, "-m", "Surface", "&"]
execute(cmd, 'os')
def thickinthehead(segmented_file,
labeled_file,
cortex_value=2,
noncortex_value=3,
labels=[],
names=[],
resize=True,
propagate=True,
output_dir='',
save_table=False,
output_table=''):
"""
Compute a simple thickness measure for each labeled cortex region volume.
Note::
- Cortex, noncortex, & label files are from the same coregistered brain.
- Calls ANTs functions: ImageMath, Threshold, ResampleImageBySpacing
- There may be slight discrepancies between volumes computed by
thickinthehead() and volumes computed by volume_for_each_label();
in 31 of 600+ ADNI 1.5T images, some volume_for_each_label() volumes
were slightly larger (in the third decimal place), presumably due to
label propagation through the cortex in thickinthehead().
This is more pronounced in ANTs vs. FreeSurfer-labeled volumes.
Example preprocessing steps ::
1. Run Freesurfer and antsCorticalThickness.sh on T1-weighted image.
2. Convert FreeSurfer volume labels (e.g., wmparc.mgz or aparc+aseg.mgz)
to cortex (2) and noncortex (3) segments using relabel_volume()
function [refer to LABELS.rst or FreeSurferColorLUT labels file].
3. Convert ANTs Atropos-segmented volume (tmpBrainSegmentation.nii.gz)
to cortex and noncortex segments, by converting 1-labels to 0 and
4-labels to 3 with the relabel_volume() function
(the latter is to include deep-gray matter with noncortical tissues).
4. Combine FreeSurfer and ANTs segmentation volumes to obtain a single
cortex (2) and noncortex (3) segmentation file using the function
combine_2labels_in_2volumes(). This function takes the union of
cortex voxels from the segmentations, the union of the noncortex
voxels from the segmentations, and overwrites intersecting cortex
and noncortex voxels with noncortex (3) labels.
ANTs tends to include more cortical gray matter at the periphery of
the brain than Freesurfer, and FreeSurfer tends to include more white
matter that extends deep into gyral folds than ANTs, so the above
attempts to remedy their differences by overlaying ANTs cortical gray
with FreeSurfer white matter.
5. Optional, see Step 2 below:
Fill segmented cortex with cortex labels and noncortex with
noncortex labels using the PropagateLabelsThroughMask() function
(which calls ImageMath ... PropagateLabelsThroughMask in ANTs).
The labels can be initialized using FreeSurfer (e.g. wmparc.mgz)
or ANTs (by applying the nonlinear inverse transform generated by
antsCorticalThickness.sh to labels in the Atropos template space).
[Note: Any further labeling steps may be applied, such as
overwriting cerebrum with intersecting cerebellum labels.]
Steps ::
1. Extract noncortex and cortex.
2. Either mask labels with cortex or fill cortex with labels.
3. Resample cortex and noncortex files from 1x1x1 to 0.5x0.5x0.5
to better represent the contours of the boundaries of the cortex.
4. Extract outer and inner boundary voxels of the cortex,
by eroding 1 (resampled) voxel for cortex voxels (2) bordering
the outside of the brain (0) and bordering noncortex (3).
5. Estimate middle cortical surface area by the average volume
of the outer and inner boundary voxels of the cortex.
6. Compute the volume of a labeled region of cortex.
7. Estimate the thickness of the labeled cortical region as the
volume of the labeled region (#6) divided by the surface area (#5).
Parameters
----------
segmented_file : string
image volume with cortex and noncortex (and any other) labels
labeled_file : string
corresponding image volume with index labels
cortex_value : integer
cortex label value in segmented_file
noncortex_value : integer
noncortex label value in segmented_file
labels : list of integers
label indices
names : list of strings
label names
resize : Boolean
resize (2x) segmented_file for more accurate thickness estimates?
propagate : Boolean
propagate labels through cortex?
output_dir : string
output directory
save_table : Boolean
save output table file with label volumes and thickness values?
output_table : string
name of output table file with label volumes and thickness values
Returns
-------
label_volume_thickness : list of lists of integers and floats
label indices, volumes, and thickness values (default -1)
output_table : string
name of output table file with label volumes and thickness values
Examples
--------
>>> from mindboggle.shapes.volume_shapes import thickinthehead
>>> segmented_file = '/Users/arno/Data/antsCorticalThickness/OASIS-TRT-20-1/antsBrainSegmentation.nii.gz'
>>> labeled_file = '/appsdir/freesurfer/subjects/OASIS-TRT-20-1/mri/labels.DKT31.manual.nii.gz'
>>> cortex_value = 2
>>> noncortex_value = 3
>>> #labels = [2]
>>> labels = range(1002,1036) + range(2002,2036)
>>> labels.remove(1004)
>>> labels.remove(2004)
>>> labels.remove(1032)
>>> labels.remove(2032)
>>> labels.remove(1033)
>>> labels.remove(2033)
>>> names = []
>>> resize = True
>>> propagate = False
>>> output_dir = ''
>>> save_table = True
>>> output_table = ''
>>> label_volume_thickness, output_table = thickinthehead(segmented_file, labeled_file, cortex_value, noncortex_value, labels, names, resize, propagate, output_dir, save_table, output_table)
"""
import os
import numpy as np
import nibabel as nb
from mindboggle.guts.utilities import execute
#-------------------------------------------------------------------------
# Output files:
#-------------------------------------------------------------------------
if output_dir:
if not os.path.exists(output_dir):
os.mkdir(output_dir)
else:
output_dir = os.getcwd()
cortex = os.path.join(output_dir, 'cortex.nii.gz')
noncortex = os.path.join(output_dir, 'noncortex.nii.gz')
temp = os.path.join(output_dir, 'temp.nii.gz')
inner_edge = os.path.join(output_dir, 'cortex_inner_edge.nii.gz')
use_outer_edge = True
if use_outer_edge:
outer_edge = os.path.join(output_dir, 'cortex_outer_edge.nii.gz')
if save_table:
if output_table:
output_table = os.path.join(os.getcwd(), output_table)
else:
output_table = os.path.join(os.getcwd(),
'thickinthehead_for_each_label.csv')
fid = open(output_table, 'w')
if names:
fid.write("name, ID, thickness (thickinthehead)\n")
else:
fid.write("ID, thickness (thickinthehead)\n")
else:
output_table = ''
#-------------------------------------------------------------------------
# Extract noncortex and cortex:
#-------------------------------------------------------------------------
cmd = [
'ThresholdImage 3', segmented_file, noncortex,
str(noncortex_value),
str(noncortex_value), '1 0'
]
execute(cmd)
cmd = [
'ThresholdImage 3', segmented_file, cortex,
str(cortex_value),
str(cortex_value), '1 0'
]
execute(cmd)
#-------------------------------------------------------------------------
# Either mask labels with cortex or fill cortex with labels:
#-------------------------------------------------------------------------
if propagate:
cmd = [
'ImageMath', '3', cortex, 'PropagateLabelsThroughMask', cortex,
labeled_file
]
execute(cmd)
else:
cmd = ['ImageMath 3', cortex, 'm', cortex, labeled_file]
execute(cmd)
#-------------------------------------------------------------------------
# Load data and dimensions:
#-------------------------------------------------------------------------
if resize:
rescale = 2.0
else:
rescale = 1.0
compute_real_volume = True
if compute_real_volume:
img = nb.load(cortex)
hdr = img.get_header()
vv_orig = np.prod(hdr.get_zooms())
vv = np.prod([x / rescale for x in hdr.get_zooms()])
cortex_data = img.get_data().ravel()
else:
vv = 1 / rescale
cortex_data = nb.load(cortex).get_data().ravel()
#-------------------------------------------------------------------------
# Resample cortex and noncortex files from 1x1x1 to 0.5x0.5x0.5
# to better represent the contours of the boundaries of the cortex:
#-------------------------------------------------------------------------
if resize:
dims = ' '.join([str(1 / rescale), str(1 / rescale), str(1 / rescale)])
cmd = ['ResampleImageBySpacing 3', cortex, cortex, dims, '0 0 1']
execute(cmd)
cmd = ['ResampleImageBySpacing 3', noncortex, noncortex, dims, '0 0 1']
execute(cmd)
#-------------------------------------------------------------------------
# Extract outer and inner boundary voxels of the cortex,
# by eroding 1 (resampled) voxel for cortex voxels (2) bordering
# the outside of the brain (0) and bordering noncortex (3):
#-------------------------------------------------------------------------
cmd = ['ImageMath 3', inner_edge, 'MD', noncortex, '1']
execute(cmd)
cmd = ['ImageMath 3', inner_edge, 'm', cortex, inner_edge]
execute(cmd)
if use_outer_edge:
cmd = ['ThresholdImage 3', cortex, outer_edge, '1 10000 1 0']
execute(cmd)
cmd = ['ImageMath 3', outer_edge, 'ME', outer_edge, '1']
execute(cmd)
cmd = ['ThresholdImage 3', outer_edge, outer_edge, '1 1 0 1']
execute(cmd)
cmd = ['ImageMath 3', outer_edge, 'm', cortex, outer_edge]
execute(cmd)
cmd = ['ThresholdImage 3', inner_edge, temp, '1 10000 1 0']
execute(cmd)
cmd = ['ThresholdImage 3', temp, temp, '1 1 0 1']
execute(cmd)
cmd = ['ImageMath 3', outer_edge, 'm', temp, outer_edge]
execute(cmd)
#-------------------------------------------------------------------------
# Load data:
#-------------------------------------------------------------------------
inner_edge_data = nb.load(inner_edge).get_data().ravel()
if use_outer_edge:
outer_edge_data = nb.load(outer_edge).get_data().ravel()
#-------------------------------------------------------------------------
# Loop through labels:
#-------------------------------------------------------------------------
if not labels:
labeled_data = nb.load(labeled_file).get_data().ravel()
labels = np.unique(labeled_data)
labels = [int(x) for x in labels]
label_volume_thickness = -1 * np.ones((len(labels), 3))
label_volume_thickness[:, 0] = labels
for ilabel, label in enumerate(labels):
if names:
name = names[ilabel]
#---------------------------------------------------------------------
# Compute thickness as a ratio of label volume and edge volume:
# - Estimate middle cortical surface area by the average volume
# of the outer and inner boundary voxels of the cortex.
# - Compute the volume of a labeled region of cortex.
# - Estimate the thickness of the labeled cortical region as the
# volume of the labeled region divided by the surface area.
#---------------------------------------------------------------------
label_cortex_volume = vv_orig * len(np.where(cortex_data == label)[0])
label_inner_edge_volume = vv * len(
np.where(inner_edge_data == label)[0])
if label_inner_edge_volume:
if use_outer_edge:
label_outer_edge_volume = \
vv * len(np.where(outer_edge_data==label)[0])
label_area = (label_inner_edge_volume +
label_outer_edge_volume) / 2.0
else:
label_area = label_inner_edge_volume
thickness = label_cortex_volume / label_area
label_volume_thickness[ilabel, 1] = label_cortex_volume
label_volume_thickness[ilabel, 2] = thickness
#print('label {0} volume: cortex={1:2.2f}, inner={2:2.2f}, '
# 'outer={3:2.2f}, area51={4:2.2f}, thickness={5:2.2f}mm'.
# format(name, label, label_cortex_volume, label_inner_edge_volume,
# label_outer_edge_volume, label_area, thickness))
if names:
print('{0} ({1}) thickness={2:2.2f}mm'.format(
name, label, thickness))
else:
print('{0}, thickness={1:2.2f}mm'.format(label, thickness))
if save_table:
if names:
fid.write('{0}, {1}, {2:2.3f}\n'.format(
name, label, thickness))
else:
fid.write('{0}, {1:2.3f}\n'.format(label, thickness))
label_volume_thickness = label_volume_thickness.transpose().tolist()
return label_volume_thickness, output_table
def thickinthehead(segmented_file, labeled_file,
cortex_value=2, noncortex_value=3, labels=[], names=[],
propagate=False, output_dir='', save_table=False,
output_table='', verbose=False):
"""
Compute a simple thickness measure for each labeled cortex region volume.
Since Mindboggle accepts FreeSurfer data as input, we include FreeSurfer
cortical thickness estimates with Mindboggle’s shape measures.
However, surface mesh reconstruction from MRI data does not always
produce favorable results. For example, we found that at least a quarter
of the over one hundred EMBARC brain images we processed through
FreeSurfer clipped ventral cortical regions, resulting in bad surface
patches in those regions. For comparison, we built this function called
thickinthehead which computes a simple thickness measure for each
cortical region using a segmentation volume rather than surfaces.
We have revised this algorithm from the original published version.
We removed upsampling to reduce memory issues for large image volumes,
and replaced the estimated volume of middle cortical layer
with an estimate of its surface area. We made these revisions to be less
susceptible to deviations in voxel size from isometric 1mm^3 voxels
for which thickinthehead was originally built.
Steps ::
1. Extract noncortex and cortex into separate files.
2. Either mask labels with cortex or fill cortex with labels.
3. Extract outer and inner boundary voxels of the cortex,
by morphologically eroding the cortex (=2) by one voxel bordering
the outside of the brain (=0) and bordering the inside of the brain
(non-cortex=3).
4. Estimate middle cortical layer's surface area by the average
surface area of the outer and inner boundary voxels of the cortex,
where surface area is roughly estimated as the average face area
of a voxel times the number of voxels.
5. Compute the volume of a labeled region of cortex.
6. Estimate the thickness of the labeled cortical region as the
volume of the labeled region (#5) divided by the
estimate of the middle cortical surface area of that region (#4).
Note::
- Cortex, noncortex, & label files are from the same coregistered brain.
- Calls ANTs functions: ImageMath and Threshold
- There may be slight discrepancies between volumes computed by
thickinthehead() and volumes computed by volume_per_label();
in 31 of 600+ ADNI 1.5T images, some volume_per_label() volumes
were slightly larger (in the third decimal place), presumably due to
label propagation through the cortex in thickinthehead().
This is more pronounced in ANTs vs. FreeSurfer-labeled volumes.
Parameters
----------
segmented_file : string
image volume with cortex and noncortex (and any other) labels
labeled_file : string
corresponding image volume with index labels
cortex_value : integer
cortex label value in segmented_file
noncortex_value : integer
noncortex label value in segmented_file
labels : list of integers
label indices
names : list of strings
label names
propagate : bool
propagate labels through cortex (or mask labels with cortex)?
output_dir : string
output directory
save_table : bool
save output table file with label volumes and thickness values?
output_table : string
name of output table file with label volumes and thickness values
verbose : bool
print statements?
Returns
-------
label_volume_thickness : list of lists of integers and floats
label indices, volumes, and thickness values (default -1)
output_table : string
name of output table file with label volumes and thickness values
Examples
--------
>>> # Example simply using ants segmentation and labels vs. hybrid segmentation:
>>> import os
>>> import numpy as np
>>> from mindboggle.shapes.volume_shapes import thickinthehead
>>> from mindboggle.mio.fetch_data import prep_tests
>>> urls, fetch_data = prep_tests()
>>> segmented_file = fetch_data(urls['ants_segmentation'], '', '.nii.gz')
>>> labeled_file = fetch_data(urls['ants_labels'], '', '.nii.gz')
>>> cortex_value = 2
>>> noncortex_value = 3
>>> #labels = [2]
>>> labels = list(range(1002,1036)) + list(range(2002,2036))
>>> labels.remove(1004)
>>> labels.remove(2004)
>>> labels.remove(1032)
>>> labels.remove(2032)
>>> labels.remove(1033)
>>> labels.remove(2033)
>>> names = []
>>> propagate = False
>>> output_dir = ''
>>> save_table = True
>>> output_table = ''
>>> verbose = False
Skip online test because it requires installation of ANTs:
>>> label_volume_thickness, output_table = thickinthehead(segmented_file,
... labeled_file, cortex_value, noncortex_value, labels, names,
... propagate, output_dir, save_table, output_table, verbose) # doctest: +SKIP
>>> [np.int("{0:.{1}f}".format(x, 5)) label_volume_thickness[0][0:10]] # doctest: +SKIP
>>> [np.float("{0:.{1}f}".format(x, 5)) for x in label_volume_thickness[1][0:5]] # doctest: +SKIP
>>> [np.float("{0:.{1}f}".format(x, 5)) for x in label_volume_thickness[2][0:5]] # doctest: +SKIP
"""
import os
import numpy as np
import nibabel as nb
from io import open
from mindboggle.guts.utilities import execute
# ------------------------------------------------------------------------
# Output files:
# ------------------------------------------------------------------------
if output_dir:
if not os.path.exists(output_dir):
os.mkdir(output_dir)
else:
output_dir = os.getcwd()
cortex = os.path.join(output_dir, 'cortex.nii.gz')
noncortex = os.path.join(output_dir, 'noncortex.nii.gz')
temp = os.path.join(output_dir, 'temp.nii.gz')
inner_edge = os.path.join(output_dir, 'cortex_inner_edge.nii.gz')
use_outer_edge = True
if use_outer_edge:
outer_edge = os.path.join(output_dir, 'cortex_outer_edge.nii.gz')
if save_table:
if output_table:
output_table = os.path.join(os.getcwd(), output_table)
else:
output_table = os.path.join(os.getcwd(),
'thickinthehead_for_each_label.csv')
fid = open(output_table, 'w')
if names:
fid.write("name, ID, thickness (thickinthehead)\n")
else:
fid.write("ID, thickness (thickinthehead)\n")
else:
output_table = ''
# ------------------------------------------------------------------------
# Extract noncortex and cortex:
# ------------------------------------------------------------------------
cmd = ['ThresholdImage', '3', segmented_file, noncortex,
str(noncortex_value), str(noncortex_value), '1 0']
execute(cmd, 'os')
cmd = ['ThresholdImage', '3', segmented_file, cortex,
str(cortex_value), str(cortex_value), '1 0']
execute(cmd, 'os')
# ------------------------------------------------------------------------
# Either mask labels with cortex or fill cortex with labels:
# ------------------------------------------------------------------------
if propagate:
cmd = ['ImageMath', '3', cortex, 'PropagateLabelsThroughMask',
cortex, labeled_file]
execute(cmd, 'os')
else:
cmd = ['ImageMath', '3', cortex, 'm', cortex, labeled_file]
execute(cmd, 'os')
# ------------------------------------------------------------------------
# Load data and dimensions:
# ------------------------------------------------------------------------
img = nb.load(cortex)
cortex_data = img.get_data().ravel()
voxsize = img.header.get_zooms()
voxvol = np.prod(voxsize)
voxarea = (voxsize[0] * voxsize[1] + \
voxsize[0] * voxsize[2] + \
voxsize[1] * voxsize[2]) / 3
# ------------------------------------------------------------------------
# Extract outer and inner boundary voxels of the cortex,
# by eroding 1 voxel for cortex voxels (=2) bordering
# the outside of the brain (=0) and bordering noncortex (=3):
# ------------------------------------------------------------------------
cmd = ['ImageMath', '3', inner_edge, 'MD', noncortex, '1']
execute(cmd, 'os')
cmd = ['ImageMath', '3', inner_edge, 'm', cortex, inner_edge]
execute(cmd, 'os')
if use_outer_edge:
cmd = ['ThresholdImage', '3', cortex, outer_edge, '1 10000 1 0']
execute(cmd, 'os')
cmd = ['ImageMath', '3', outer_edge, 'ME', outer_edge, '1']
execute(cmd, 'os')
cmd = ['ThresholdImage', '3', outer_edge, outer_edge, '1 1 0 1']
execute(cmd, 'os')
cmd = ['ImageMath', '3', outer_edge, 'm', cortex, outer_edge]
execute(cmd, 'os')
cmd = ['ThresholdImage', '3', inner_edge, temp, '1 10000 1 0']
execute(cmd, 'os')
cmd = ['ThresholdImage', '3', temp, temp, '1 1 0 1']
execute(cmd, 'os')
cmd = ['ImageMath', '3', outer_edge, 'm', temp, outer_edge]
execute(cmd, 'os')
# ------------------------------------------------------------------------
# Load data:
# ------------------------------------------------------------------------
inner_edge_data = nb.load(inner_edge).get_data().ravel()
if use_outer_edge:
outer_edge_data = nb.load(outer_edge).get_data().ravel()
# ------------------------------------------------------------------------
# Loop through labels:
# ------------------------------------------------------------------------
if not labels:
labeled_data = nb.load(labeled_file).get_data().ravel()
labels = np.unique(labeled_data)
labels = [int(x) for x in labels]
label_volume_thickness = -1 * np.ones((len(labels), 3))
label_volume_thickness[:, 0] = labels
for ilabel, label in enumerate(labels):
if names:
name = names[ilabel]
# --------------------------------------------------------------------
# Compute thickness as a ratio of label volume and layer surface area:
# - Estimate middle cortical surface area by the average area
# of the outer and inner boundary voxels of the cortex.
# - Surface area is roughly estimated as the average face area
# of a voxel times the number of voxels.
# - Compute the volume of a labeled region of cortex.
# - Estimate the thickness of the labeled cortical region as the
# volume of the labeled region divided by the middle surface area.
# --------------------------------------------------------------------
label_cortex_volume = voxvol * len(np.where(cortex_data==label)[0])
label_inner_edge_area = voxarea * \
len(np.where(inner_edge_data==label)[0])
if label_inner_edge_area:
if use_outer_edge:
label_outer_edge_area = \
voxarea * len(np.where(outer_edge_data==label)[0])
label_area = (label_inner_edge_area +
label_outer_edge_area) / 2.0
else:
label_area = label_inner_edge_area
thickness = label_cortex_volume / label_area
label_volume_thickness[ilabel, 1] = label_cortex_volume
label_volume_thickness[ilabel, 2] = thickness
if save_table:
if names:
if verbose:
print('{0} ({1}) thickinthehead thickness = '
'{2:2.2f}mm'.format(name, label, thickness))
fid.write('{0}, {1}, {2:2.3f}\n'.format(name, label,
thickness))
else:
if verbose:
print('{0} thickinthehead thickness = {1:2.2f}mm'.
format(label, thickness))
fid.write('{0}, {1:2.3f}\n'.format(label, thickness))
label_volume_thickness = label_volume_thickness.transpose().tolist()
return label_volume_thickness, output_table
def ImageMath(volume1, volume2, operator='m', output_file=''):
"""
Use the ImageMath function in ANTs to perform operation on two volumes::
m : Multiply --- use vm for vector multiply
+ : Add --- use v+ for vector add
- : Subtract --- use v- for vector subtract
/ : Divide
^ : Power
exp : Take exponent exp(imagevalue*value)
addtozero : add image-b to image-a only over points where image-a has zero values
overadd : replace image-a pixel with image-b pixel if image-b pixel is non-zero
abs : absolute value
total : Sums up values in an image or in image1*image2 (img2 is the probability mask)
mean : Average of values in an image or in image1*image2 (img2 is the probability mask)
vtotal : Sums up volumetrically weighted values in an image or in image1*image2 (img2 is the probability mask)
Decision : Computes result=1./(1.+exp(-1.0*( pix1-0.25)/pix2))
Neg : Produce image negative
Parameters
----------
volume1 : string
nibabel-readable image volume
volume2 : string
nibabel-readable image volume
operator : string
ImageMath string corresponding to mathematical operator
output_file : string
nibabel-readable image volume
Returns
-------
output_file : string
name of output nibabel-readable image volume
Examples
--------
>>> # Mask head with brain mask:
>>> import os
>>> from mindboggle.thirdparty.ants import ImageMath
>>> from mindboggle.mio.fetch_data import prep_tests
>>> urls, fetch_data = prep_tests()
>>> volume1 = fetch_data(urls['T1_001'])
>>> volume2 = fetch_data(urls['ants_mask'])
>>> operator = 'm'
>>> output_file = ''
>>> output_file = ImageMath(volume1, volume2, operator, output_file) # doctest: +SKIP
View result (skip test):
>>> from mindboggle.mio.plots import plot_volumes
>>> plot_volumes(output_file) # doctest: +SKIP
"""
import os
from mindboggle.guts.utilities import execute
if not output_file:
output_file = os.path.join(os.getcwd(),
os.path.basename(volume1) + '_' +
os.path.basename(volume2))
cmd = ['ImageMath', '3', output_file, operator, volume1, volume2]
execute(cmd, 'os')
if not os.path.exists(output_file):
raise IOError("ImageMath did not create " + output_file + ".")
return output_file
def antsApplyTransformsToPoints(points, transform_files, inverse_booleans=[0]):
"""
Run ANTs antsApplyTransformsToPoints function to transform points.
(Creates pre- and post-transformed .csv points files for ANTs.)
Parameters
----------
points : list of lists of three integers
point coordinate data
transform_files : list
transform file names
inverse_booleans : list
for each transform, one to apply inverse of transform (otherwise zero)
Returns
-------
transformed_points : list of lists of three integers
transformed point coordinate data
Examples
--------
>>> from mindboggle.thirdparty.ants import antsApplyTransformsToPoints
>>> from mindboggle.mio.vtks import read_vtk
>>> transform_files = ['/Users/arno/Data/antsCorticalThickness/Twins-2-1/antsTemplateToSubject1GenericAffine.mat','/Users/arno/Data/antsCorticalThickness/Twins-2-1/antsTemplateToSubject0Warp.nii.gz','/Users/arno/Data/antsCorticalThickness/Twins-2-1/antsSubjectToTemplate0GenericAffine.mat','/Users/arno/Data/antsCorticalThickness/Twins-2-1/antsSubjectToTemplate1Warp.nii.gz']
>>> transform_files = [transform_files[0],transform_files[1],'/Users/arno/Data/mindboggle_cache/f36e3d5d99f7c4a9bb70e2494ed7340b/OASIS-30_Atropos_template_to_MNI152_affine.txt']
>>> vtk_file = '/Users/arno/mindboggle_working/Twins-2-1/Mindboggle/_hemi_lh/Surface_to_vtk/lh.pial.vtk'
>>> points, indices, lines, faces, scalars, scalar_names, npoints, input_vtk = read_vtk(vtk_file)
>>> inverse_booleans = [0,0,1]
>>> transformed_points = antsApplyTransformsToPoints(points, transform_files, inverse_booleans)
"""
import os
from mindboggle.guts.utilities import execute
#-------------------------------------------------------------------------
# Write points (x,y,z,1) to a .csv file:
#-------------------------------------------------------------------------
points_file = os.path.join(os.getcwd(), 'points.csv')
fid = open(points_file, 'wa')
fid.write('x,y,z,t\n')
for point in points:
string_of_zeros = (4 - len(point)) * ',0'
fid.write(','.join([str(x) for x in point]) + string_of_zeros + '\n')
fid.close()
#-------------------------------------------------------------------------
# Apply transforms to points in .csv file:
#-------------------------------------------------------------------------
transformed_points_file = os.path.join(os.getcwd(),
'transformed_points.csv')
transform_string = ''
for ixfm, transform_file in enumerate(transform_files):
transform_string += " --t [{0},{1}]".\
format(transform_file, str(inverse_booleans[ixfm]))
cmd = ['antsApplyTransformsToPoints', '-d', '3', '-i', points_file,
'-o', transformed_points_file, transform_string]
execute(cmd, 'os')
if not os.path.exists(transformed_points_file):
str1 = "antsApplyTransformsToPoints did not create "
raise(IOError(str1 + transformed_points_file + "."))
#-------------------------------------------------------------------------
# Return transformed points:
#-------------------------------------------------------------------------
fid = open(transformed_points_file, 'r')
lines = fid.readlines()
fid.close()
transformed_points = []
for iline, line in enumerate(lines):
if iline > 0:
point_xyz1 = [float(x) for x in line.split(',')]
transformed_points.append(point_xyz1[0:3])
return transformed_points
def PropagateLabelsThroughMask(mask, labels, mask_index=None,
output_file='', binarize=True, stopvalue=''):
"""
Use ANTs to fill a binary volume mask with initial labels.
This program uses ThresholdImage and the ImageMath
PropagateLabelsThroughMask functions in ANTs.
ThresholdImage ImageDimension ImageIn.ext outImage.ext
threshlo threshhi <insideValue> <outsideValue>
PropagateLabelsThroughMask: Final output is the propagated label image.
ImageMath ImageDimension Out.ext PropagateLabelsThroughMask
speed/binaryimagemask.nii.gz initiallabelimage.nii.gz ...
Parameters
----------
mask : string
nibabel-readable image volume
labels : string
nibabel-readable image volume with integer labels
mask_index : integer (optional)
mask with just voxels having this value
output_file : string
nibabel-readable labeled image volume
binarize : Boolean
binarize mask?
stopvalue : integer
stopping value
Returns
-------
output_file : string
name of labeled output nibabel-readable image volume
Examples
--------
>>> import os
>>> from mindboggle.thirdparty.ants import PropagateLabelsThroughMask
>>> from mindboggle.mio.plots import plot_volumes
>>> path = os.path.join(os.environ['MINDBOGGLE_DATA'])
>>> labels = os.path.join(path, 'arno', 'labels', 'labels.DKT25.manual.nii.gz')
>>> mask = os.path.join(path, 'arno', 'mri', 't1weighted_brain.nii.gz')
>>> mask_index = None
>>> output_file = ''
>>> binarize = True
>>> stopvalue = None
>>> output_file = PropagateLabelsThroughMask(mask, labels, mask_index, output_file, binarize, stopvalue)
>>> # View
>>> plot_volumes(output_file)
"""
import os
from mindboggle.guts.utilities import execute
if not output_file:
#output_file = os.path.join(os.getcwd(),
# 'PropagateLabelsThroughMask.nii.gz')
output_file = os.path.join(os.getcwd(),
os.path.basename(labels) + '_through_' +
os.path.basename(mask))
print('mask: {0}, labels: {1}'.format(mask, labels))
# Binarize image volume:
if binarize:
temp_file = os.path.join(os.getcwd(),
'PropagateLabelsThroughMask.nii.gz')
cmd = ['ThresholdImage', '3', mask, temp_file, '0 1 0 1']
execute(cmd, 'os')
mask = temp_file
# Mask with just voxels having mask_index value:
if mask_index:
mask2 = os.path.join(os.getcwd(), 'temp.nii.gz')
cmd = 'ThresholdImage 3 {0} {1} {2} {3} 1 0'.format(mask, mask2,
mask_index, mask_index)
execute(cmd)
else:
mask2 = mask
# Propagate labels:
cmd = ['ImageMath', '3', output_file, 'PropagateLabelsThroughMask',
mask2, labels]
if stopvalue:
cmd.extend(stopvalue)
execute(cmd, 'os')
if not os.path.exists(output_file):
raise(IOError("ImageMath did not create " + output_file + "."))
return output_file
def thickinthehead(segmented_file,
labeled_file,
cortex_value=2,
noncortex_value=3,
labels=[],
names=[],
resize=True,
propagate=True,
output_dir='',
save_table=False,
output_table='',
verbose=False):
"""
Compute a simple thickness measure for each labeled cortex region volume.
Since Mindboggle accepts FreeSurfer data as input, we include FreeSurfer
cortical thickness estimates with Mindboggle’s shape measures.
However, surface mesh reconstruction from MRI data does not always
produce favorable results. For example, we found that at least a quarter
of the over one hundred EMBARC brain images we processed through
FreeSurfer clipped ventral cortical regions, resulting in bad surface
patches in those regions. For comparison, we built a function called
thickinthehead which computes a simple thickness measure for each
cortical region using the hybrid segmentation volume rather than surfaces.
The thickinthehead function first saves a brain volume that has been
segmented into cortex and non-cortex voxels into separate binary files,
then resamples these cortex and non-cortex files from, for example,
1mm^3 to 0.5mm^3 voxel dimensions to better represent the contours
of the cortex, then extracts outer and inner boundary voxels of the cortex
by morphologically eroding the cortex by one (resampled) voxel bordering
the outside of the brain and bordering the inside of the brain
(non-cortex). Then it estimates the middle cortical surface area by the
average volume of the outer and inner boundary voxels of the cortex.
Finally, it estimates the thickness of a labeled cortical region as the
volume of the labeled region divided by the surface area of that region.
We compared thickinthehead and FreeSurfer cortical thickness estimates
for 16 cortical regions in 40 EMBARC control subjects (unpublished
results) with published estimates based on manual delineations of MR
images (Kabani, 2001). Forty percent of FreeSurfer estimates for the 640
labels were in the range of the published values, whereas almost ninety
percent of thickinthehead’s estimates were within range. ANTs values
deviated further from the published estimates and were less reliable
(greater inter-subject ranges) than the FreeSurfer or thickinthehead
values.
Note::
- Cortex, noncortex, & label files are from the same coregistered brain.
- Calls ANTs functions: ImageMath, Threshold, ResampleImageBySpacing
- There may be slight discrepancies between volumes computed by
thickinthehead() and volumes computed by volume_per_label();
in 31 of 600+ ADNI 1.5T images, some volume_per_label() volumes
were slightly larger (in the third decimal place), presumably due to
label propagation through the cortex in thickinthehead().
This is more pronounced in ANTs vs. FreeSurfer-labeled volumes.
Example preprocessing steps ::
1. Run Freesurfer and antsCorticalThickness.sh on T1-weighted image.
2. Convert FreeSurfer volume labels (e.g., wmparc.mgz or aparc+aseg.mgz)
to cortex (2) and noncortex (3) segments using relabel_volume()
function [refer to labels.rst or FreeSurferColorLUT labels file].
3. Convert ANTs Atropos-segmented volume (tmpBrainSegmentation.nii.gz)
to cortex and noncortex segments, by converting 1-labels to 0 and
4-labels to 3 with the relabel_volume() function
(the latter is to include deep-gray matter with noncortical tissues).
4. Combine FreeSurfer and ANTs segmentation volumes to obtain a single
cortex (2) and noncortex (3) segmentation file using the function
combine_2labels_in_2volumes(). This function takes the union of
cortex voxels from the segmentations, the union of the noncortex
voxels from the segmentations, and overwrites intersecting cortex
and noncortex voxels with noncortex (3) labels.
ANTs tends to include more cortical gray matter at the periphery of
the brain than Freesurfer, and FreeSurfer tends to include more white
matter that extends deep into gyral folds than ANTs, so the above
attempts to remedy their differences by overlaying ANTs cortical gray
with FreeSurfer white matter.
5. Optional, see Step 2 below:
Fill segmented cortex with cortex labels and noncortex with
noncortex labels using the PropagateLabelsThroughMask() function
(which calls ImageMath ... PropagateLabelsThroughMask in ANTs).
The labels can be initialized using FreeSurfer (e.g. wmparc.mgz)
or ANTs (by applying the nonlinear inverse transform generated by
antsCorticalThickness.sh to labels in the Atropos template space).
[Note: Any further labeling steps may be applied, such as
overwriting cerebrum with intersecting cerebellum labels.]
Steps ::
1. Extract noncortex and cortex.
2. Either mask labels with cortex or fill cortex with labels.
3. Resample cortex and noncortex files from 1x1x1 to 0.5x0.5x0.5
to better represent the contours of the boundaries of the cortex.
4. Extract outer and inner boundary voxels of the cortex,
by eroding 1 (resampled) voxel for cortex voxels (2) bordering
the outside of the brain (0) and bordering noncortex (3).
5. Estimate middle cortical surface area by the average volume
of the outer and inner boundary voxels of the cortex.
6. Compute the volume of a labeled region of cortex.
7. Estimate the thickness of the labeled cortical region as the
volume of the labeled region (#6) divided by the surface area (#5).
Parameters
----------
segmented_file : string
image volume with cortex and noncortex (and any other) labels
labeled_file : string
corresponding image volume with index labels
cortex_value : integer
cortex label value in segmented_file
noncortex_value : integer
noncortex label value in segmented_file
labels : list of integers
label indices
names : list of strings
label names
resize : bool
resize (2x) segmented_file for more accurate thickness estimates?
propagate : bool
propagate labels through cortex?
output_dir : string
output directory
save_table : bool
save output table file with label volumes and thickness values?
output_table : string
name of output table file with label volumes and thickness values
verbose : bool
print statements?
Returns
-------
label_volume_thickness : list of lists of integers and floats
label indices, volumes, and thickness values (default -1)
output_table : string
name of output table file with label volumes and thickness values
Examples
--------
>>> # Example simply using ants segmentation and labels:
>>> import os
>>> import numpy as np
>>> from mindboggle.shapes.volume_shapes import thickinthehead
>>> from mindboggle.mio.fetch_data import prep_tests
>>> urls, fetch_data = prep_tests()
>>> segmented_file = fetch_data(urls['ants_segmentation'], '', '.nii.gz')
>>> labeled_file = fetch_data(urls['ants_labels'], '', '.nii.gz')
>>> cortex_value = 2
>>> noncortex_value = 3
>>> #labels = [2]
>>> labels = list(range(1002,1036)) + list(range(2002,2036))
>>> labels.remove(1004)
>>> labels.remove(2004)
>>> labels.remove(1032)
>>> labels.remove(2032)
>>> labels.remove(1033)
>>> labels.remove(2033)
>>> names = []
>>> resize = True
>>> propagate = False
>>> output_dir = ''
>>> save_table = True
>>> output_table = ''
>>> verbose = False
Skip online test because it requires installation of ANTs:
>>> label_volume_thickness, output_table = thickinthehead(segmented_file,
... labeled_file, cortex_value, noncortex_value, labels, names,
... resize, propagate, output_dir, save_table, output_table, verbose) # doctest: +SKIP
>>> [np.int("{0:.{1}f}".format(x, 5)) label_volume_thickness[0][0:10]] # doctest: +SKIP
[1002, 1003, 1005, 1006, 1007, 1008, 1009, 1010, 1011 1012]
>>> [np.float("{0:.{1}f}".format(x, 5)) for x in label_volume_thickness[1][0:5]] # doctest: +SKIP
[3136.99383, 7206.98582, 3257.99359, 1950.99616, 12458.97549]
>>> [np.float("{0:.{1}f}".format(x, 5)) for x in label_volume_thickness[2][0:5]] # doctest: +SKIP
[3.8639, 3.69637, 2.56334, 4.09336, 4.52592]
"""
import os
import numpy as np
import nibabel as nb
from io import open
from mindboggle.guts.utilities import execute
# ------------------------------------------------------------------------
# Output files:
# ------------------------------------------------------------------------
if output_dir:
if not os.path.exists(output_dir):
os.mkdir(output_dir)
else:
output_dir = os.getcwd()
cortex = os.path.join(output_dir, 'cortex.nii.gz')
noncortex = os.path.join(output_dir, 'noncortex.nii.gz')
temp = os.path.join(output_dir, 'temp.nii.gz')
inner_edge = os.path.join(output_dir, 'cortex_inner_edge.nii.gz')
use_outer_edge = True
if use_outer_edge:
outer_edge = os.path.join(output_dir, 'cortex_outer_edge.nii.gz')
if save_table:
if output_table:
output_table = os.path.join(os.getcwd(), output_table)
else:
output_table = os.path.join(os.getcwd(),
'thickinthehead_for_each_label.csv')
fid = open(output_table, 'w')
if names:
fid.write("name, ID, thickness (thickinthehead)\n")
else:
fid.write("ID, thickness (thickinthehead)\n")
else:
output_table = ''
# ------------------------------------------------------------------------
# ants command paths:
# ------------------------------------------------------------------------
ants_thresh = 'ThresholdImage'
ants_math = 'ImageMath'
ants_resample = 'ResampleImageBySpacing'
# ------------------------------------------------------------------------
# Extract noncortex and cortex:
# ------------------------------------------------------------------------
cmd = [
ants_thresh, '3', segmented_file, noncortex,
str(noncortex_value),
str(noncortex_value), '1 0'
]
execute(cmd, 'os')
cmd = [
ants_thresh, '3', segmented_file, cortex,
str(cortex_value),
str(cortex_value), '1 0'
]
execute(cmd, 'os')
# ------------------------------------------------------------------------
# Either mask labels with cortex or fill cortex with labels:
# ------------------------------------------------------------------------
if propagate:
cmd = [
ants_math, '3', cortex, 'PropagateLabelsThroughMask', cortex,
labeled_file
]
execute(cmd, 'os')
else:
cmd = [ants_math, '3', cortex, 'm', cortex, labeled_file]
execute(cmd, 'os')
# ------------------------------------------------------------------------
# Load data and dimensions:
# ------------------------------------------------------------------------
if resize:
rescale = 2.0
else:
rescale = 1.0
compute_real_volume = True
if compute_real_volume:
img = nb.load(cortex)
hdr = img.get_header()
vv_orig = np.prod(hdr.get_zooms())
vv = np.prod([x / rescale for x in hdr.get_zooms()])
cortex_data = img.get_data().ravel()
else:
vv = 1 / rescale
cortex_data = nb.load(cortex).get_data().ravel()
# ------------------------------------------------------------------------
# Resample cortex and noncortex files from 1x1x1 to 0.5x0.5x0.5
# to better represent the contours of the boundaries of the cortex:
# ------------------------------------------------------------------------
if resize:
dims = ' '.join([str(1 / rescale), str(1 / rescale), str(1 / rescale)])
cmd = [ants_resample, '3', cortex, cortex, dims, '0 0 1']
execute(cmd, 'os')
cmd = [ants_resample, '3', noncortex, noncortex, dims, '0 0 1']
execute(cmd, 'os')
# ------------------------------------------------------------------------
# Extract outer and inner boundary voxels of the cortex,
# by eroding 1 (resampled) voxel for cortex voxels (2) bordering
# the outside of the brain (0) and bordering noncortex (3):
# ------------------------------------------------------------------------
cmd = [ants_math, '3', inner_edge, 'MD', noncortex, '1']
execute(cmd, 'os')
cmd = [ants_math, '3', inner_edge, 'm', cortex, inner_edge]
execute(cmd, 'os')
if use_outer_edge:
cmd = [ants_thresh, '3', cortex, outer_edge, '1 10000 1 0']
execute(cmd, 'os')
cmd = [ants_math, '3', outer_edge, 'ME', outer_edge, '1']
execute(cmd, 'os')
cmd = [ants_thresh, '3', outer_edge, outer_edge, '1 1 0 1']
execute(cmd, 'os')
cmd = [ants_math, '3', outer_edge, 'm', cortex, outer_edge]
execute(cmd, 'os')
cmd = [ants_thresh, '3', inner_edge, temp, '1 10000 1 0']
execute(cmd, 'os')
cmd = [ants_thresh, '3', temp, temp, '1 1 0 1']
execute(cmd, 'os')
cmd = [ants_math, '3', outer_edge, 'm', temp, outer_edge]
execute(cmd, 'os')
# ------------------------------------------------------------------------
# Load data:
# ------------------------------------------------------------------------
inner_edge_data = nb.load(inner_edge).get_data().ravel()
if use_outer_edge:
outer_edge_data = nb.load(outer_edge).get_data().ravel()
# ------------------------------------------------------------------------
# Loop through labels:
# ------------------------------------------------------------------------
if not labels:
labeled_data = nb.load(labeled_file).get_data().ravel()
labels = np.unique(labeled_data)
labels = [int(x) for x in labels]
label_volume_thickness = -1 * np.ones((len(labels), 3))
label_volume_thickness[:, 0] = labels
for ilabel, label in enumerate(labels):
if names:
name = names[ilabel]
# --------------------------------------------------------------------
# Compute thickness as a ratio of label volume and edge volume:
# - Estimate middle cortical surface area by the average volume
# of the outer and inner boundary voxels of the cortex.
# - Compute the volume of a labeled region of cortex.
# - Estimate the thickness of the labeled cortical region as the
# volume of the labeled region divided by the surface area.
# --------------------------------------------------------------------
label_cortex_volume = vv_orig * len(np.where(cortex_data == label)[0])
label_inner_edge_volume = vv * len(
np.where(inner_edge_data == label)[0])
if label_inner_edge_volume:
if use_outer_edge:
label_outer_edge_volume = \
vv * len(np.where(outer_edge_data==label)[0])
label_area = (label_inner_edge_volume +
label_outer_edge_volume) / 2.0
else:
label_area = label_inner_edge_volume
thickness = label_cortex_volume / label_area
label_volume_thickness[ilabel, 1] = label_cortex_volume
label_volume_thickness[ilabel, 2] = thickness
if save_table:
if names:
if verbose:
print('{0} ({1}) thickinthehead thickness = '
'{2:2.2f}mm'.format(name, label, thickness))
fid.write('{0}, {1}, {2:2.3f}\n'.format(
name, label, thickness))
else:
if verbose:
print(
'{0} thickinthehead thickness = {1:2.2f}mm'.format(
label, thickness))
fid.write('{0}, {1:2.3f}\n'.format(label, thickness))
label_volume_thickness = label_volume_thickness.transpose().tolist()
return label_volume_thickness, output_table
def plot_mask_surface(vtk_file,
mask_file='',
nonmask_value=-1,
masked_output='',
remove_nonmask=False,
program='vtkviewer',
use_colormap=False,
colormap_file='',
background_value=-1):
"""
Use vtkviewer or mayavi2 to visualize VTK surface mesh data.
If a mask_file is provided, a temporary masked file is saved,
and it is this file that is viewed.
If using vtkviewer, optionally provide colormap file
or set $COLORMAP environment variable.
Parameters
----------
vtk_file : string
name of VTK surface mesh file
mask_file : string
name of VTK surface mesh file to mask vtk_file vertices
nonmask_value : integer
nonmask (usually background) value
masked_output : string
temporary masked output file name
remove_nonmask : bool
remove vertices that are not in mask? (otherwise assign nonmask_value)
program : string {'vtkviewer', 'mayavi2'}
program to visualize VTK file
use_colormap : bool
use Paraview-style XML colormap file set by $COLORMAP env variable?
colormap_file : string
use colormap in given file if use_colormap==True? if empty and
use_colormap==True, use file set by $COLORMAP environment variable
background_value : integer or float
background value
Examples
--------
>>> import os
>>> from mindboggle.mio.plots import plot_mask_surface
>>> from mindboggle.mio.fetch_data import prep_tests
>>> urls, fetch_data = prep_tests()
>>> vtk_file = fetch_data(urls['freesurfer_labels'], '', '.vtk')
>>> os.rename(vtk_file, vtk_file + '.nii.gz')
>>> vtk_file = vtk_file + '.nii.gz'
>>> mask_file = ''
>>> nonmask_value = 0 #-1
>>> masked_output = ''
>>> remove_nonmask = True
>>> program = 'vtkviewer'
>>> use_colormap = True
>>> colormap_file = ''
>>> background_value = -1
>>> plot_mask_surface(vtk_file, mask_file, nonmask_value, masked_output,
... remove_nonmask, program, use_colormap, colormap_file,
... background_value) # doctest: +SKIP
"""
import os
import numpy as np
from mindboggle.guts.mesh import keep_faces, reindex_faces_points
from mindboggle.guts.utilities import execute
from mindboggle.mio.plots import plot_surfaces
from mindboggle.mio.vtks import read_scalars, rewrite_scalars, \
read_vtk, write_vtk
# ------------------------------------------------------------------------
# Filter mesh with non-background values from a second (same-size) mesh:
# ------------------------------------------------------------------------
if mask_file:
mask, name = read_scalars(mask_file, True, True)
if not masked_output:
masked_output = os.path.join(os.getcwd(), 'temp.vtk')
file_to_plot = masked_output
# --------------------------------------------------------------------
# Remove nonmask-valued vertices:
# --------------------------------------------------------------------
if remove_nonmask:
# ----------------------------------------------------------------
# Load VTK files:
# ----------------------------------------------------------------
points, indices, lines, faces, scalars, scalar_names, npoints, \
input_vtk = read_vtk(vtk_file, True, True)
# ----------------------------------------------------------------
# Find mask indices, remove nonmask faces, and reindex:
# ----------------------------------------------------------------
Imask = [i for i, x in enumerate(mask) if x != nonmask_value]
mask_faces = keep_faces(faces, Imask)
mask_faces, points, \
original_indices = reindex_faces_points(mask_faces, points)
# ----------------------------------------------------------------
# Write VTK file with scalar values:
# ----------------------------------------------------------------
if np.ndim(scalars) == 1:
scalar_type = type(scalars[0]).__name__
elif np.ndim(scalars) == 2:
scalar_type = type(scalars[0][0]).__name__
else:
print("Undefined scalar type!")
write_vtk(file_to_plot,
points, [], [],
mask_faces,
scalars[original_indices].tolist(),
scalar_names,
scalar_type=scalar_type)
else:
scalars, name = read_scalars(vtk_file, True, True)
scalars[mask == nonmask_value] = nonmask_value
rewrite_scalars(vtk_file, file_to_plot, scalars, ['scalars'], [],
background_value)
else:
file_to_plot = vtk_file
# ------------------------------------------------------------------------
# Display with vtkviewer.py:
# ------------------------------------------------------------------------
if program == 'vtkviewer':
plot_surfaces(file_to_plot,
use_colormap=use_colormap,
colormap_file=colormap_file)
# ------------------------------------------------------------------------
# Display with mayavi2:
# ------------------------------------------------------------------------
elif program == 'mayavi2':
cmd = ["mayavi2", "-d", file_to_plot, "-m", "Surface", "&"]
execute(cmd, 'os')
def antsApplyTransformsToPoints(points, transform_files, inverse_booleans=[0]):
"""
Run ANTs antsApplyTransformsToPoints function to transform points.
(Creates pre- and post-transformed .csv points files for ANTs.)
Parameters
----------
points : list of lists of three integers
point coordinate data
transform_files : list
transform file names
inverse_booleans : list
for each transform, one to apply inverse of transform (otherwise zero)
Returns
-------
transformed_points : list of lists of three integers
transformed point coordinate data
Examples
--------
>>> from mindboggle.thirdparty.ants import antsApplyTransformsToPoints
>>> from mindboggle.mio.vtks import read_vtk
>>> transform_files = ['/Users/arno/Data/antsCorticalThickness/Twins-2-1/antsTemplateToSubject1GenericAffine.mat','/Users/arno/Data/antsCorticalThickness/Twins-2-1/antsTemplateToSubject0Warp.nii.gz','/Users/arno/Data/antsCorticalThickness/Twins-2-1/antsSubjectToTemplate0GenericAffine.mat','/Users/arno/Data/antsCorticalThickness/Twins-2-1/antsSubjectToTemplate1Warp.nii.gz']
>>> transform_files = [transform_files[0],transform_files[1],'/Users/arno/Data/mindboggle_cache/f36e3d5d99f7c4a9bb70e2494ed7340b/OASIS-30_Atropos_template_to_MNI152_affine.txt']
>>> vtk_file = '/Users/arno/mindboggle_working/Twins-2-1/Mindboggle/_hemi_lh/Surface_to_vtk/lh.pial.vtk'
>>> faces, lines, indices, points, npoints, scalars, name, foo1 = read_vtk(vtk_file)
>>> inverse_booleans = [0,0,1]
>>> transformed_points = antsApplyTransformsToPoints(points, transform_files, inverse_booleans)
"""
import os
from mindboggle.guts.utilities import execute
#-------------------------------------------------------------------------
# Write points (x,y,z,1) to a .csv file:
#-------------------------------------------------------------------------
points_file = os.path.join(os.getcwd(), 'points.csv')
fid = open(points_file, 'wa')
fid.write('x,y,z,t\n')
for point in points:
string_of_zeros = (4 - len(point)) * ',0'
fid.write(','.join([str(x) for x in point]) + string_of_zeros + '\n')
fid.close()
#-------------------------------------------------------------------------
# Apply transforms to points in .csv file:
#-------------------------------------------------------------------------
transformed_points_file = os.path.join(os.getcwd(),
'transformed_points.csv')
transform_string = ''
for ixfm, transform_file in enumerate(transform_files):
transform_string += " --t [{0},{1}]".\
format(transform_file, str(inverse_booleans[ixfm]))
cmd = [
'antsApplyTransformsToPoints', '-d', '3', '-i', points_file, '-o',
transformed_points_file, transform_string
]
execute(cmd, 'os')
if not os.path.exists(transformed_points_file):
str1 = "antsApplyTransformsToPoints did not create "
raise (IOError(str1 + transformed_points_file + "."))
#-------------------------------------------------------------------------
# Return transformed points:
#-------------------------------------------------------------------------
fid = open(transformed_points_file, 'r')
lines = fid.readlines()
fid.close()
transformed_points = []
for iline, line in enumerate(lines):
if iline > 0:
point_xyz1 = [float(x) for x in line.split(',')]
transformed_points.append(point_xyz1[0:3])
return transformed_points
def thickinthehead(segmented_file, labeled_file, cortex_value=2,
noncortex_value=3, labels=[], names=[], resize=True,
propagate=True, output_dir='', save_table=False,
output_table='', ants_path='', verbose=False):
"""
Compute a simple thickness measure for each labeled cortex region volume.
Note::
- Cortex, noncortex, & label files are from the same coregistered brain.
- Calls ANTs functions: ImageMath, Threshold, ResampleImageBySpacing
- There may be slight discrepancies between volumes computed by
thickinthehead() and volumes computed by volume_for_each_label();
in 31 of 600+ ADNI 1.5T images, some volume_for_each_label() volumes
were slightly larger (in the third decimal place), presumably due to
label propagation through the cortex in thickinthehead().
This is more pronounced in ANTs vs. FreeSurfer-labeled volumes.
Example preprocessing steps ::
1. Run Freesurfer and antsCorticalThickness.sh on T1-weighted image.
2. Convert FreeSurfer volume labels (e.g., wmparc.mgz or aparc+aseg.mgz)
to cortex (2) and noncortex (3) segments using relabel_volume()
function [refer to LABELS.rst or FreeSurferColorLUT labels file].
3. Convert ANTs Atropos-segmented volume (tmpBrainSegmentation.nii.gz)
to cortex and noncortex segments, by converting 1-labels to 0 and
4-labels to 3 with the relabel_volume() function
(the latter is to include deep-gray matter with noncortical tissues).
4. Combine FreeSurfer and ANTs segmentation volumes to obtain a single
cortex (2) and noncortex (3) segmentation file using the function
combine_2labels_in_2volumes(). This function takes the union of
cortex voxels from the segmentations, the union of the noncortex
voxels from the segmentations, and overwrites intersecting cortex
and noncortex voxels with noncortex (3) labels.
ANTs tends to include more cortical gray matter at the periphery of
the brain than Freesurfer, and FreeSurfer tends to include more white
matter that extends deep into gyral folds than ANTs, so the above
attempts to remedy their differences by overlaying ANTs cortical gray
with FreeSurfer white matter.
5. Optional, see Step 2 below:
Fill segmented cortex with cortex labels and noncortex with
noncortex labels using the PropagateLabelsThroughMask() function
(which calls ImageMath ... PropagateLabelsThroughMask in ANTs).
The labels can be initialized using FreeSurfer (e.g. wmparc.mgz)
or ANTs (by applying the nonlinear inverse transform generated by
antsCorticalThickness.sh to labels in the Atropos template space).
[Note: Any further labeling steps may be applied, such as
overwriting cerebrum with intersecting cerebellum labels.]
Steps ::
1. Extract noncortex and cortex.
2. Either mask labels with cortex or fill cortex with labels.
3. Resample cortex and noncortex files from 1x1x1 to 0.5x0.5x0.5
to better represent the contours of the boundaries of the cortex.
4. Extract outer and inner boundary voxels of the cortex,
by eroding 1 (resampled) voxel for cortex voxels (2) bordering
the outside of the brain (0) and bordering noncortex (3).
5. Estimate middle cortical surface area by the average volume
of the outer and inner boundary voxels of the cortex.
6. Compute the volume of a labeled region of cortex.
7. Estimate the thickness of the labeled cortical region as the
volume of the labeled region (#6) divided by the surface area (#5).
Parameters
----------
segmented_file : string
image volume with cortex and noncortex (and any other) labels
labeled_file : string
corresponding image volume with index labels
cortex_value : integer
cortex label value in segmented_file
noncortex_value : integer
noncortex label value in segmented_file
labels : list of integers
label indices
names : list of strings
label names
resize : bool
resize (2x) segmented_file for more accurate thickness estimates?
propagate : bool
propagate labels through cortex?
output_dir : string
output directory
save_table : bool
save output table file with label volumes and thickness values?
output_table : string
name of output table file with label volumes and thickness values
verbose : bool
print statements?
Returns
-------
label_volume_thickness : list of lists of integers and floats
label indices, volumes, and thickness values (default -1)
output_table : string
name of output table file with label volumes and thickness values
Examples
--------
>>> # Example simply using ants segmentation and labels:
>>> import os
>>> import numpy as np
>>> from mindboggle.shapes.volume_shapes import thickinthehead
>>> from mindboggle.mio.fetch_data import prep_tests
>>> urls, fetch_data = prep_tests()
>>> segmented_file = fetch_data(urls['ants_segmentation'])
>>> labeled_file = fetch_data(urls['ants_labels'])
>>> cortex_value = 2
>>> noncortex_value = 3
>>> #labels = [2]
>>> labels = list(range(1002,1036)) + list(range(2002,2036))
>>> labels.remove(1004)
>>> labels.remove(2004)
>>> labels.remove(1032)
>>> labels.remove(2032)
>>> labels.remove(1033)
>>> labels.remove(2033)
>>> names = []
>>> resize = True
>>> propagate = False
>>> output_dir = ''
>>> save_table = True
>>> output_table = ''
>>> verbose = False
>>> label_volume_thickness, output_table = thickinthehead(segmented_file,
... labeled_file, cortex_value, noncortex_value, labels, names,
... resize, propagate, output_dir, save_table, output_table, verbose)
>>> print(np.array_str(np.array(label_volume_thickness[0][0:10]),
... precision=5, suppress_small=True))
[ 1002. 1003. 1005. 1006. 1007. 1008. 1009. 1010. 1011. 1012.]
>>> print(np.array_str(np.array(label_volume_thickness[1][0:5]),
... precision=5, suppress_small=True))
[ 3136.99383 7206.98582 3257.99359 1950.99616 12458.97549]
>>> print(np.array_str(np.array(label_volume_thickness[2][0:5]),
... precision=5, suppress_small=True))
[ 3.8639 3.69637 2.56334 4.09336 4.52592]
"""
import os
import numpy as np
import nibabel as nb
from io import open
from mindboggle.guts.utilities import execute
#-------------------------------------------------------------------------
# Output files:
#-------------------------------------------------------------------------
if output_dir:
if not os.path.exists(output_dir):
os.mkdir(output_dir)
else:
output_dir = os.getcwd()
cortex = os.path.join(output_dir, 'cortex.nii.gz')
noncortex = os.path.join(output_dir, 'noncortex.nii.gz')
temp = os.path.join(output_dir, 'temp.nii.gz')
inner_edge = os.path.join(output_dir, 'cortex_inner_edge.nii.gz')
use_outer_edge = True
if use_outer_edge:
outer_edge = os.path.join(output_dir, 'cortex_outer_edge.nii.gz')
if save_table:
if output_table:
output_table = os.path.join(os.getcwd(), output_table)
else:
output_table = os.path.join(os.getcwd(),
'thickinthehead_for_each_label.csv')
fid = open(output_table, 'w')
if names:
fid.write("name, ID, thickness (thickinthehead)\n")
else:
fid.write("ID, thickness (thickinthehead)\n")
else:
output_table = ''
#-------------------------------------------------------------------------
# ants command paths:
#-------------------------------------------------------------------------
ants_thresh = 'ThresholdImage'
ants_math = 'ImageMath'
ants_resample = 'ResampleImageBySpacing'
#-------------------------------------------------------------------------
# Extract noncortex and cortex:
#-------------------------------------------------------------------------
cmd = [ants_thresh, '3', segmented_file, noncortex,
str(noncortex_value), str(noncortex_value), '1 0']
execute(cmd, 'os')
cmd = [ants_thresh, '3', segmented_file, cortex,
str(cortex_value), str(cortex_value), '1 0']
execute(cmd, 'os')
#-------------------------------------------------------------------------
# Either mask labels with cortex or fill cortex with labels:
#-------------------------------------------------------------------------
if propagate:
cmd = [ants_math, '3', cortex, 'PropagateLabelsThroughMask',
cortex, labeled_file]
execute(cmd, 'os')
else:
cmd = [ants_math, '3', cortex, 'm', cortex, labeled_file]
execute(cmd, 'os')
#-------------------------------------------------------------------------
# Load data and dimensions:
#-------------------------------------------------------------------------
if resize:
rescale = 2.0
else:
rescale = 1.0
compute_real_volume = True
if compute_real_volume:
img = nb.load(cortex)
hdr = img.get_header()
vv_orig = np.prod(hdr.get_zooms())
vv = np.prod([x/rescale for x in hdr.get_zooms()])
cortex_data = img.get_data().ravel()
else:
vv = 1/rescale
cortex_data = nb.load(cortex).get_data().ravel()
#-------------------------------------------------------------------------
# Resample cortex and noncortex files from 1x1x1 to 0.5x0.5x0.5
# to better represent the contours of the boundaries of the cortex:
#-------------------------------------------------------------------------
if resize:
dims = ' '.join([str(1/rescale), str(1/rescale), str(1/rescale)])
cmd = [ants_resample, '3', cortex, cortex, dims, '0 0 1']
execute(cmd, 'os')
cmd = [ants_resample, '3', noncortex, noncortex, dims, '0 0 1']
execute(cmd, 'os')
#-------------------------------------------------------------------------
# Extract outer and inner boundary voxels of the cortex,
# by eroding 1 (resampled) voxel for cortex voxels (2) bordering
# the outside of the brain (0) and bordering noncortex (3):
#-------------------------------------------------------------------------
cmd = [ants_math, '3', inner_edge, 'MD', noncortex, '1']
execute(cmd, 'os')
cmd = [ants_math, '3', inner_edge, 'm', cortex, inner_edge]
execute(cmd, 'os')
if use_outer_edge:
cmd = [ants_thresh, '3', cortex, outer_edge, '1 10000 1 0']
execute(cmd, 'os')
cmd = [ants_math, '3', outer_edge, 'ME', outer_edge, '1']
execute(cmd, 'os')
cmd = [ants_thresh, '3', outer_edge, outer_edge, '1 1 0 1']
execute(cmd, 'os')
cmd = [ants_math, '3', outer_edge, 'm', cortex, outer_edge]
execute(cmd, 'os')
cmd = [ants_thresh, '3', inner_edge, temp, '1 10000 1 0']
execute(cmd, 'os')
cmd = [ants_thresh, '3', temp, temp, '1 1 0 1']
execute(cmd, 'os')
cmd = [ants_math, '3', outer_edge, 'm', temp, outer_edge]
execute(cmd, 'os')
#-------------------------------------------------------------------------
# Load data:
#-------------------------------------------------------------------------
inner_edge_data = nb.load(inner_edge).get_data().ravel()
if use_outer_edge:
outer_edge_data = nb.load(outer_edge).get_data().ravel()
#-------------------------------------------------------------------------
# Loop through labels:
#-------------------------------------------------------------------------
if not labels:
labeled_data = nb.load(labeled_file).get_data().ravel()
labels = np.unique(labeled_data)
labels = [int(x) for x in labels]
label_volume_thickness = -1 * np.ones((len(labels), 3))
label_volume_thickness[:, 0] = labels
for ilabel, label in enumerate(labels):
if names:
name = names[ilabel]
#---------------------------------------------------------------------
# Compute thickness as a ratio of label volume and edge volume:
# - Estimate middle cortical surface area by the average volume
# of the outer and inner boundary voxels of the cortex.
# - Compute the volume of a labeled region of cortex.
# - Estimate the thickness of the labeled cortical region as the
# volume of the labeled region divided by the surface area.
#---------------------------------------------------------------------
label_cortex_volume = vv_orig * len(np.where(cortex_data==label)[0])
label_inner_edge_volume = vv * len(np.where(inner_edge_data==label)[0])
if label_inner_edge_volume:
if use_outer_edge:
label_outer_edge_volume = \
vv * len(np.where(outer_edge_data==label)[0])
label_area = (label_inner_edge_volume +
label_outer_edge_volume) / 2.0
else:
label_area = label_inner_edge_volume
thickness = label_cortex_volume / label_area
label_volume_thickness[ilabel, 1] = label_cortex_volume
label_volume_thickness[ilabel, 2] = thickness
if save_table:
if names:
if verbose:
print('{0} ({1}) thickinthehead thickness = '
'{2:2.2f}mm'.format(name, label, thickness))
fid.write('{0}, {1}, {2:2.3f}\n'.format(name, label,
thickness))
else:
if verbose:
print('{0} thickinthehead thickness = {1:2.2f}mm'.
format(label, thickness))
fid.write('{0}, {1:2.3f}\n'.format(label, thickness))
label_volume_thickness = label_volume_thickness.transpose().tolist()
return label_volume_thickness, output_table
def antsApplyTransformsToPoints(points, transform_files,
inverse_booleans=[0]):
"""
Run ANTs antsApplyTransformsToPoints function to transform points.
(Creates pre- and post-transformed .csv points files for ANTs.)
Parameters
----------
points : list of lists of three integers
point coordinate data
transform_files : list
transform file names
inverse_booleans : list
for each transform, one to apply inverse of transform (otherwise zero)
Returns
-------
transformed_points : list of lists of three integers
transformed point coordinate data
Examples
--------
>>> import os
>>> import numpy as np
>>> from mindboggle.thirdparty.ants import antsApplyTransformsToPoints
>>> from mindboggle.mio.vtks import read_points
>>> from mindboggle.mio.fetch_data import prep_tests
>>> urls, fetch_data = prep_tests()
>>> xfm1 = fetch_data(urls['ants_affine_template2subject'])
>>> xfm2 = fetch_data(urls['ants_warp_template2subject'])
>>> xfm3 = fetch_data(urls['OASIS-30_Atropos_template_to_MNI152_affine'])
>>> transform_files = [xfm1, xfm2, xfm3]
>>> vtk_file = fetch_data(urls['left_pial'])
>>> points = read_points(vtk_file)
>>> inverse_booleans = [0,0,1]
>>> transformed_points = antsApplyTransformsToPoints(points,
... transform_files, inverse_booleans) # doctest: +SKIP
>>> print(np.array_str(np.array(transformed_points[0:5]),
... precision=5, suppress_small=True)) # doctest: +SKIP
[[-11.23189 -46.78223 -39.88869]
[-11.71384 -46.87075 -40.13328]
[-12.56237 -46.99126 -40.04564]
[ -9.66693 -46.0446 -41.36334]
[-10.67998 -46.45458 -40.7572 ]]
"""
import os
from io import open
from mindboggle.guts.utilities import execute
#-------------------------------------------------------------------------
# Write points (x,y,z,1) to a .csv file:
#-------------------------------------------------------------------------
points_file = os.path.join(os.getcwd(), 'points.csv')
fid = open(points_file, 'wa')
fid.write('x,y,z,t\n')
for point in points:
string_of_zeros = (4 - len(point)) * ',0'
fid.write(','.join([str(x) for x in point]) + string_of_zeros + '\n')
fid.close()
#-------------------------------------------------------------------------
# Apply transforms to points in .csv file:
#-------------------------------------------------------------------------
transformed_points_file = os.path.join(os.getcwd(),
'transformed_points.csv')
transform_string = ''
for ixfm, transform_file in enumerate(transform_files):
transform_string += " --t [{0},{1}]".\
format(transform_file, str(inverse_booleans[ixfm]))
cmd = ['antsApplyTransformsToPoints', '-d', '3', '-i', points_file,
'-o', transformed_points_file, transform_string]
try:
execute(cmd, 'os')
except:
raise Exception("Cannot find antsApplyTransformsToPoints command.")
if not os.path.exists(transformed_points_file):
raise IOError("antsApplyTransformsToPoints did not create {0}.".
format(transformed_points_file))
#-------------------------------------------------------------------------
# Return transformed points:
#-------------------------------------------------------------------------
fid = open(transformed_points_file, 'rb')
lines = fid.readlines()
fid.close()
transformed_points = []
for iline, line in enumerate(lines):
if iline > 0:
point_xyz1 = [float(x) for x in line.split(',')]
transformed_points.append(point_xyz1[0:3])
return transformed_points
def thickinthehead(segmented_file, labeled_file,
cortex_value=2, noncortex_value=3, labels=[], names=[],
propagate=False, output_dir='', save_table=False,
output_table='', verbose=False):
"""
Compute a simple thickness measure for each labeled cortex region volume.
Since Mindboggle accepts FreeSurfer data as input, we include FreeSurfer
cortical thickness estimates with Mindboggle’s shape measures.
However, surface mesh reconstruction from MRI data does not always
produce favorable results. For example, we found that at least a quarter
of the over one hundred EMBARC brain images we processed through
FreeSurfer clipped ventral cortical regions, resulting in bad surface
patches in those regions. For comparison, we built this function called
thickinthehead which computes a simple thickness measure for each
cortical region using a segmentation volume rather than surfaces.
We have revised this algorithm from the original published version.
We removed upsampling to reduce memory issues for large image volumes,
and replaced the estimated volume of middle cortical layer
with an estimate of its surface area. We made these revisions to be less
susceptible to deviations in voxel size from isometric 1mm^3 voxels
for which thickinthehead was originally built.
Steps ::
1. Extract noncortex and cortex into separate files.
2. Either mask labels with cortex or fill cortex with labels.
3. Extract outer and inner boundary voxels of the cortex,
by morphologically eroding the cortex (=2) by one voxel bordering
the outside of the brain (=0) and bordering the inside of the brain
(non-cortex=3).
4. Estimate middle cortical layer's surface area by the average
surface area of the outer and inner boundary voxels of the cortex,
where surface area is roughly estimated as the average face area
of a voxel times the number of voxels.
5. Compute the volume of a labeled region of cortex.
6. Estimate the thickness of the labeled cortical region as the
volume of the labeled region (#5) divided by the
estimate of the middle cortical surface area of that region (#4).
Note::
- Cortex, noncortex, & label files are from the same coregistered brain.
- Calls ANTs functions: ImageMath and Threshold
- There may be slight discrepancies between volumes computed by
thickinthehead() and volumes computed by volume_per_label();
in 31 of 600+ ADNI 1.5T images, some volume_per_label() volumes
were slightly larger (in the third decimal place), presumably due to
label propagation through the cortex in thickinthehead().
This is more pronounced in ANTs vs. FreeSurfer-labeled volumes.
Parameters
----------
segmented_file : string
image volume with cortex and noncortex (and any other) labels
labeled_file : string
corresponding image volume with index labels
cortex_value : integer
cortex label value in segmented_file
noncortex_value : integer
noncortex label value in segmented_file
labels : list of integers
label indices
names : list of strings
label names
propagate : bool
propagate labels through cortex (or mask labels with cortex)?
output_dir : string
output directory
save_table : bool
save output table file with label volumes and thickness values?
output_table : string
name of output table file with label volumes and thickness values
verbose : bool
print statements?
Returns
-------
label_volume_thickness : list of lists of integers and floats
label indices, volumes, and thickness values (default -1)
output_table : string
name of output table file with label volumes and thickness values
Examples
--------
>>> # Example simply using ants segmentation and labels vs. hybrid segmentation:
>>> import os
>>> import numpy as np
>>> from mindboggle.shapes.volume_shapes import thickinthehead
>>> from mindboggle.mio.fetch_data import prep_tests
>>> urls, fetch_data = prep_tests()
>>> segmented_file = fetch_data(urls['ants_segmentation'], '', '.nii.gz')
>>> labeled_file = fetch_data(urls['ants_labels'], '', '.nii.gz')
>>> cortex_value = 2
>>> noncortex_value = 3
>>> #labels = [2]
>>> labels = list(range(1002,1036)) + list(range(2002,2036))
>>> labels.remove(1004)
>>> labels.remove(2004)
>>> labels.remove(1032)
>>> labels.remove(2032)
>>> labels.remove(1033)
>>> labels.remove(2033)
>>> names = []
>>> propagate = False
>>> output_dir = ''
>>> save_table = True
>>> output_table = ''
>>> verbose = False
Skip online test because it requires installation of ANTs:
>>> label_volume_thickness, output_table = thickinthehead(segmented_file,
... labeled_file, cortex_value, noncortex_value, labels, names,
... propagate, output_dir, save_table, output_table, verbose) # doctest: +SKIP
>>> [np.int("{0:.{1}f}".format(x, 5)) label_volume_thickness[0][0:10]] # doctest: +SKIP
>>> [np.float("{0:.{1}f}".format(x, 5)) for x in label_volume_thickness[1][0:5]] # doctest: +SKIP
>>> [np.float("{0:.{1}f}".format(x, 5)) for x in label_volume_thickness[2][0:5]] # doctest: +SKIP
"""
import os
import numpy as np
import nibabel as nb
from io import open
from mindboggle.guts.utilities import execute
# ------------------------------------------------------------------------
# Output files:
# ------------------------------------------------------------------------
if output_dir:
if not os.path.exists(output_dir):
os.mkdir(output_dir)
else:
output_dir = os.getcwd()
cortex = os.path.join(output_dir, 'cortex.nii.gz')
noncortex = os.path.join(output_dir, 'noncortex.nii.gz')
temp = os.path.join(output_dir, 'temp.nii.gz')
inner_edge = os.path.join(output_dir, 'cortex_inner_edge.nii.gz')
use_outer_edge = True
if use_outer_edge:
outer_edge = os.path.join(output_dir, 'cortex_outer_edge.nii.gz')
if save_table:
if output_table:
output_table = os.path.join(os.getcwd(), output_table)
else:
output_table = os.path.join(os.getcwd(),
'thickinthehead_for_each_label.csv')
fid = open(output_table, 'w', encoding='utf-8')
if names:
fid.write("name, ID, thickness (thickinthehead)\n")
else:
fid.write("ID, thickness (thickinthehead)\n")
else:
output_table = ''
# ------------------------------------------------------------------------
# Extract noncortex and cortex:
# ------------------------------------------------------------------------
cmd = ['ThresholdImage', '3', segmented_file, noncortex,
str(noncortex_value), str(noncortex_value), '1 0']
execute(cmd, 'os')
cmd = ['ThresholdImage', '3', segmented_file, cortex,
str(cortex_value), str(cortex_value), '1 0']
execute(cmd, 'os')
# ------------------------------------------------------------------------
# Either mask labels with cortex or fill cortex with labels:
# ------------------------------------------------------------------------
if propagate:
cmd = ['ImageMath', '3', cortex, 'PropagateLabelsThroughMask',
cortex, labeled_file]
execute(cmd, 'os')
else:
cmd = ['ImageMath', '3', cortex, 'm', cortex, labeled_file]
execute(cmd, 'os')
# ------------------------------------------------------------------------
# Load data and dimensions:
# ------------------------------------------------------------------------
img = nb.load(cortex)
cortex_data = img.get_data().ravel()
voxsize = img.header.get_zooms()
voxvol = np.prod(voxsize)
voxarea = (voxsize[0] * voxsize[1] + \
voxsize[0] * voxsize[2] + \
voxsize[1] * voxsize[2]) / 3
# ------------------------------------------------------------------------
# Extract outer and inner boundary voxels of the cortex,
# by eroding 1 voxel for cortex voxels (=2) bordering
# the outside of the brain (=0) and bordering noncortex (=3):
# ------------------------------------------------------------------------
cmd = ['ImageMath', '3', inner_edge, 'MD', noncortex, '1']
execute(cmd, 'os')
cmd = ['ImageMath', '3', inner_edge, 'm', cortex, inner_edge]
execute(cmd, 'os')
if use_outer_edge:
cmd = ['ThresholdImage', '3', cortex, outer_edge, '1 10000 1 0']
execute(cmd, 'os')
cmd = ['ImageMath', '3', outer_edge, 'ME', outer_edge, '1']
execute(cmd, 'os')
cmd = ['ThresholdImage', '3', outer_edge, outer_edge, '1 1 0 1']
execute(cmd, 'os')
cmd = ['ImageMath', '3', outer_edge, 'm', cortex, outer_edge]
execute(cmd, 'os')
cmd = ['ThresholdImage', '3', inner_edge, temp, '1 10000 1 0']
execute(cmd, 'os')
cmd = ['ThresholdImage', '3', temp, temp, '1 1 0 1']
execute(cmd, 'os')
cmd = ['ImageMath', '3', outer_edge, 'm', temp, outer_edge]
execute(cmd, 'os')
# ------------------------------------------------------------------------
# Load data:
# ------------------------------------------------------------------------
inner_edge_data = nb.load(inner_edge).get_data().ravel()
if use_outer_edge:
outer_edge_data = nb.load(outer_edge).get_data().ravel()
# ------------------------------------------------------------------------
# Loop through labels:
# ------------------------------------------------------------------------
if not labels:
labeled_data = nb.load(labeled_file).get_data().ravel()
labels = np.unique(labeled_data)
labels = [int(x) for x in labels]
label_volume_thickness = -1 * np.ones((len(labels), 3))
label_volume_thickness[:, 0] = labels
for ilabel, label in enumerate(labels):
if names:
name = names[ilabel]
# --------------------------------------------------------------------
# Compute thickness as a ratio of label volume and layer surface area:
# - Estimate middle cortical surface area by the average area
# of the outer and inner boundary voxels of the cortex.
# - Surface area is roughly estimated as the average face area
# of a voxel times the number of voxels.
# - Compute the volume of a labeled region of cortex.
# - Estimate the thickness of the labeled cortical region as the
# volume of the labeled region divided by the middle surface area.
# --------------------------------------------------------------------
label_cortex_volume = voxvol * len(np.where(cortex_data==label)[0])
label_inner_edge_area = voxarea * \
len(np.where(inner_edge_data==label)[0])
if label_inner_edge_area:
if use_outer_edge:
label_outer_edge_area = \
voxarea * len(np.where(outer_edge_data==label)[0])
label_area = (label_inner_edge_area +
label_outer_edge_area) / 2.0
else:
label_area = label_inner_edge_area
thickness = label_cortex_volume / label_area
label_volume_thickness[ilabel, 1] = label_cortex_volume
label_volume_thickness[ilabel, 2] = thickness
if save_table:
if names:
if verbose:
print('{0} ({1}) thickinthehead thickness = '
'{2:2.2f}mm'.format(name, label, thickness))
fid.write('{0}, {1}, {2:2.3f}\n'.format(name, label,
thickness))
else:
if verbose:
print('{0} thickinthehead thickness = {1:2.2f}mm'.
format(label, thickness))
fid.write('{0}, {1:2.3f}\n'.format(label, thickness))
label_volume_thickness = label_volume_thickness.transpose().tolist()
return label_volume_thickness, output_table
def PropagateLabelsThroughMask(mask, labels, mask_index=None, output_file='',
binarize=True, stopvalue=''):
"""
Use ANTs to fill a binary volume mask with initial labels.
This program uses ThresholdImage and the ImageMath
PropagateLabelsThroughMask functions in ANTs.
ThresholdImage ImageDimension ImageIn.ext outImage.ext
threshlo threshhi <insideValue> <outsideValue>
PropagateLabelsThroughMask: Final output is the propagated label image.
ImageMath ImageDimension Out.ext PropagateLabelsThroughMask
speed/binaryimagemask.nii.gz initiallabelimage.nii.gz ...
Parameters
----------
mask : string
nibabel-readable image volume
labels : string
nibabel-readable image volume with integer labels
mask_index : integer (optional)
mask with just voxels having this value
output_file : string (optional)
nibabel-readable labeled image volume
binarize : bool (optional)
binarize mask?
stopvalue : integer (optional)
stopping value
Returns
-------
output_file : string
name of labeled output nibabel-readable image volume
Examples
--------
>>> # Propagate FreeSurfer labels through brain mask:
>>> import os
>>> from mindboggle.thirdparty.ants import PropagateLabelsThroughMask
>>> from mindboggle.mio.fetch_data import prep_tests
>>> urls, fetch_data = prep_tests()
>>> labels = fetch_data(urls['freesurfer_labels'])
>>> mask = fetch_data(urls['ants_mask'])
>>> mask_index = None
>>> output_file = ''
>>> binarize = True
>>> stopvalue = None
>>> output_file = PropagateLabelsThroughMask(mask, labels, mask_index,
... output_file, binarize, stopvalue) # doctest: +SKIP
View result (skip test):
>>> from mindboggle.mio.plots import plot_volumes
>>> plot_volumes(output_file) # doctest: +SKIP
"""
import os
from mindboggle.guts.utilities import execute
if not output_file:
#output_file = os.path.join(os.getcwd(),
# 'PropagateLabelsThroughMask.nii.gz')
output_file = os.path.join(os.getcwd(),
os.path.basename(labels) + '_through_' +
os.path.basename(mask))
print('mask: {0}, labels: {1}'.format(mask, labels))
# Binarize image volume:
if binarize:
temp_file = os.path.join(os.getcwd(),
'PropagateLabelsThroughMask.nii.gz')
cmd = ['ThresholdImage', '3', mask, temp_file, '0 1 0 1']
execute(cmd, 'os')
mask = temp_file
# Mask with just voxels having mask_index value:
if mask_index:
mask2 = os.path.join(os.getcwd(), 'temp.nii.gz')
cmd = ['ThresholdImage', '3', mask, mask2,
str(mask_index), str(mask_index)]
execute(cmd, 'os')
else:
mask2 = mask
# Propagate labels:
cmd = ['ImageMath', '3', output_file, 'PropagateLabelsThroughMask',
mask2, labels]
if stopvalue:
cmd.extend(str(stopvalue))
execute(cmd, 'os')
if not os.path.exists(output_file):
raise IOError("ImageMath did not create " + output_file + ".")
return output_file