def get_wm_mask(wm=data_path + 'SUB1_wm_mask.nii.gz', resample=None):
"""
Get me a white matter mask. Resample if need be
"""
bm_nii = ni.load(wm)
if resample is not None:
return ozv.resample_volume(wm, resample).get_data()
else:
return ni.load(wm).get_data()
def get_wm_mask(wm=data_path + 'SUB1_wm_mask.nii.gz', resample=None):
"""
Get me a white matter mask. Resample if need be
"""
bm_nii = ni.load(wm)
if resample is not None:
return ozv.resample_volume(wm, resample).get_data()
else:
return ni.load(wm).get_data()
def get_brain_mask(bm=data_path + 'brainMask.nii.gz',resample=None):
"""
Get the brain mask (derived from DWI using mrDiffusion) and resample if
needed
"""
bm_nii = ni.load(bm)
if resample is not None:
return ozv.resample_volume(bm, resample).get_data()
else:
return ni.load(bm).get_data()
def get_brain_mask(bm=data_path + 'brainMask.nii.gz', resample=None):
"""
Get the brain mask (derived from DWI using mrDiffusion) and resample if
needed
"""
bm_nii = ni.load(bm)
if resample is not None:
return ozv.resample_volume(bm, resample).get_data()
else:
return ni.load(bm).get_data()
def get_t1(subject, resample=None):
"""
Get the high-res T1-weighted anatomical scan. If requested, resample it to
the resolution of a nifti file for which the name is provided as input
"""
t1=data_path + '/%s/%s_t1.nii.gz'%(subject, subject)
t1_nii = ni.load(t1)
if resample is not None:
return ozv.resample_volume(t1, resample).get_data()
else:
return ni.load(t1).get_data()
def get_t1(subject, resample=None):
"""
Get the high-res T1-weighted anatomical scan. If requested, resample it to
the resolution of a nifti file for which the name is provided as input
"""
t1 = data_path + '/%s/%s_t1.nii.gz' % (subject, subject)
t1_nii = ni.load(t1)
if resample is not None:
return ozv.resample_volume(t1, resample).get_data()
else:
return ni.load(t1).get_data()
def test_resample_volume():
"""
Testing resampling of one volume into another volumes space (a t1 into a
dwi space, for example)
"""
source = ni.Nifti1Image(np.random.randn(10,10,10), np.eye(4))
target = ni.Nifti1Image(np.random.randn(10,10,10), np.eye(4))
new_vol = ozv.resample_volume(source, target)
npt.assert_equal(new_vol.shape, target.shape[:3])
def make_wm_mask(seg_path, dwi_path, out_path=data_path + 'wm_mask.nii.gz',
n_IQR=2):
"""
Function that makes a white-matter mask from an itk-gray class file and DWI
data, at the resolution of the DWI data.
Parameters
----------
seg_path : str
Full path to an itk-gray segmentation
dwi_path : str
Full path to some DWI data (used to determine the mean diffusivity in
each voxel, see below)
out_path : str (optional)
Where to put the resulting file
n_IQR: float (optional)
How many IQRs away for the definition of MD outliers (see below)
Note
----
This follows [Jeurissen2012]_. First, we resample a segmentation based on a
T1-weighted image. Then, we calculate the mean diffusivity in a
coregistered DWI image. We then also exclude from the segmentation white
matter voxels that have
MD > median (MD) + n_IQR * IQR (MD)
Where IQR is the interquartile range. Note that per default we take a
slightly more restrictive criterion (of 2 * IQR, instead of 1.5 * IQR),
based on some empirical looking at data. Not so sure why there would be a
difference.
[Jeurissen2012]Jeurissen B, Leemans A, Tournier, J-D, Jones, DK and Sijbers
J (2012). Investigating the prevalence of complex fiber configurations in
white matter tissue with diffusion magnetic resonance imaging. Human Brain
Mapping. doi: 10.1002/hbm.2209
"""
dwi_nii, dwi_bvecs, dwi_bvals = make_data_set(dwi_path)
# Resample the segmentation image to the DWI resolution:
seg_resamp = ozv.resample_volume(seg_path, dwi_nii)
# Find WM voxels (this is ITK gray coding 3=LHWM, 4=RHWM)
wm_idx = np.where(np.logical_or(seg_resamp==3, seg_resamp==4))
vol = np.zeros(seg_resamp.shape)
vol[wm_idx] = 1
# OK - now we need to find and exclude MD outliers:
TM = ozm.TensorModel(dwi_nii, dwi_bvecs, dwi_bvals, mask=vol,
params_file='temp')
MD = TM.mean_diffusivity
IQR = (stats.scoreatpercentile(MD[np.isfinite(MD)],75) -
stats.scoreatpercentile(MD[np.isfinite(MD)],25))
cutoff = np.median(MD[np.isfinite(MD)]) + n_IQR * IQR
cutoff_idx = np.where(MD > cutoff)
# Null 'em out
vol[cutoff_idx] = 0
# Now, let's save some output:
ni.Nifti1Image(vol, dwi_ni.get_affine()).to_filename(out_path)
def make_wm_mask(seg_path,
dwi_path,
out_path=data_path + 'wm_mask.nii.gz',
n_IQR=2):
"""
Function that makes a white-matter mask from an itk-gray class file and DWI
data, at the resolution of the DWI data.
Parameters
----------
seg_path : str
Full path to an itk-gray segmentation
dwi_path : str
Full path to some DWI data (used to determine the mean diffusivity in
each voxel, see below)
out_path : str (optional)
Where to put the resulting file
n_IQR: float (optional)
How many IQRs away for the definition of MD outliers (see below)
Note
----
This follows [Jeurissen2012]_. First, we resample a segmentation based on a
T1-weighted image. Then, we calculate the mean diffusivity in a
coregistered DWI image. We then also exclude from the segmentation white
matter voxels that have
MD > median (MD) + n_IQR * IQR (MD)
Where IQR is the interquartile range. Note that per default we take a
slightly more restrictive criterion (of 2 * IQR, instead of 1.5 * IQR),
based on some empirical looking at data. Not so sure why there would be a
difference.
[Jeurissen2012]Jeurissen B, Leemans A, Tournier, J-D, Jones, DK and Sijbers
J (2012). Investigating the prevalence of complex fiber configurations in
white matter tissue with diffusion magnetic resonance imaging. Human Brain
Mapping. doi: 10.1002/hbm.2209
"""
dwi_nii, dwi_bvecs, dwi_bvals = make_data_set(dwi_path)
# Resample the segmentation image to the DWI resolution:
seg_resamp = ozv.resample_volume(seg_path, dwi_nii)
# Find WM voxels (this is ITK gray coding 3=LHWM, 4=RHWM)
wm_idx = np.where(np.logical_or(seg_resamp == 3, seg_resamp == 4))
vol = np.zeros(seg_resamp.shape)
vol[wm_idx] = 1
# OK - now we need to find and exclude MD outliers:
TM = ozm.TensorModel(dwi_nii,
dwi_bvecs,
dwi_bvals,
mask=vol,
params_file='temp')
MD = TM.mean_diffusivity
IQR = (stats.scoreatpercentile(MD[np.isfinite(MD)], 75) -
stats.scoreatpercentile(MD[np.isfinite(MD)], 25))
cutoff = np.median(MD[np.isfinite(MD)]) + n_IQR * IQR
cutoff_idx = np.where(MD > cutoff)
# Null 'em out
vol[cutoff_idx] = 0
# Now, let's save some output:
ni.Nifti1Image(vol, dwi_ni.get_affine()).to_filename(out_path)
