def calc_compcor_components(data_filename, num_components, mask_filename):
if num_components < 1:
raise ValueError(
'Improper value for num_components ({0}), should be >= 1.'.format(
num_components))
try:
image_data = nb.load(data_filename).get_data().astype(np.float64)
except:
print('Unable to load data from {0}'.format(data_filename))
raise
try:
binary_mask = nb.load(mask_filename).get_data().astype(np.int16)
except:
print('Unable to load data from {0}'.format(mask_filename))
if not safe_shape(image_data, binary_mask):
raise ValueError(
'The data in {0} and {1} do not have a consistent shape'.format(
data_filename, mask_filename))
# make sure that the values in binary_mask are binary
binary_mask[binary_mask > 0] = 1
binary_mask[binary_mask != 1] = 0
# reduce the image data to only the voxels in the binary mask
image_data = image_data[binary_mask == 1, :]
# filter out any voxels whose variance equals 0
print('Removing zero variance components')
image_data = image_data[image_data.std(1) != 0, :]
if image_data.shape.count(0):
err = "\n\n[!] No wm or csf signals left after removing those " \
"with zero variance.\n\n"
raise Exception(err)
print('Detrending and centering data')
Y = signal.detrend(image_data, axis=1, type='linear').T
Yc = Y - np.tile(Y.mean(0), (Y.shape[0], 1))
Yc = Yc / np.tile(
np.array(Yc.std(0)).reshape(1, Yc.shape[1]), (Yc.shape[0], 1))
print('Calculating SVD decomposition of Y*Y\'')
U, S, Vh = np.linalg.svd(Yc, full_matrices=False)
# write out the resulting regressor file
regressor_file = os.path.join(os.getcwd(), 'compcor_regressors.1D')
np.savetxt(regressor_file,
U[:, :num_components],
delimiter='\t',
fmt='%16g')
return regressor_file
def extract_tissue_data(data_file,
ho_mask_file,
wm_seg_file, csf_seg_file, gm_seg_file,
wm_threshold=0.0, csf_threshold=0.0, gm_threshold=0.0):
import numpy as np
import nibabel as nb
import os
from CPAC.nuisance import erode_mask
from CPAC.utils import safe_shape
data = nb.load(data_file).get_data().astype('float64')
ho_mask = nb.load(ho_mask_file).get_data().astype('float64')
wm_seg = nb.load(wm_seg_file).get_data().astype('float64')
csf_seg = nb.load(csf_seg_file).get_data().astype('float64')
gm_seg = nb.load(gm_seg_file).get_data().astype('float64')
print 'Tissues extraction thresholds wm %d, csf %d, gm %d' % (wm_threshold,
csf_threshold,
gm_threshold)
if not safe_shape(data, ho_mask, wm_seg, csf_seg, gm_seg):
raise ValueError('Spatial dimensions for data, masks and tissues do not match')
wm_mask = erode_mask(wm_seg > wm_threshold)
# Only take the CSF at the lateral ventricals as labled in the Harvard
# Oxford parcellation regions 4 and 43
csf_mask = (csf_seg > csf_threshold)*((ho_mask==43) + (ho_mask == 4))
gm_mask = erode_mask(gm_seg > gm_threshold)
wm_sigs = data[wm_mask]
csf_sigs = data[csf_mask]
gm_sigs = data[gm_mask]
file_wm = os.path.join(os.getcwd(), 'wm_signals.npy')
file_csf = os.path.join(os.getcwd(), 'csf_signals.npy')
file_gm = os.path.join(os.getcwd(), 'gm_signals.npy')
np.save(file_wm, wm_sigs)
np.save(file_csf, csf_sigs)
np.save(file_gm, gm_sigs)
nii = nb.load(wm_seg_file)
wm_mask_file = os.path.join(os.getcwd(), 'wm_mask.nii.gz')
csf_mask_file = os.path.join(os.getcwd(), 'csf_mask.nii.gz')
gm_mask_file = os.path.join(os.getcwd(), 'gm_mask.nii.gz')
nb.Nifti1Image(wm_mask, header=nii.get_header(), affine=nii.get_affine()).to_filename(wm_mask_file)
nb.Nifti1Image(csf_mask, header=nii.get_header(), affine=nii.get_affine()).to_filename(csf_mask_file)
nb.Nifti1Image(gm_mask, header=nii.get_header(), affine=nii.get_affine()).to_filename(gm_mask_file)
return file_wm, file_csf, file_gm
def nifti_individual_stability(subject_file, roi_mask_file, n_bootstraps, k_clusters, cbb_block_size = None, affinity_threshold = 0.5):
"""
Calculate the individual stability matrix for a single subject by using Circular Block Bootstrapping method
for time-series data.
Parameters
----------
subject_file : string
Nifti file of a subject
roi_mask_file : string
Region of interest (this method is too computationally intensive to perform on a whole-brain volume)
n_bootstraps : integer
Number of bootstraps
k_clusters : integer
Number of clusters
cbb_block_size : integer, optional
Size of the time-series block when performing circular block bootstrap
affinity_threshold : float, optional
Minimum threshold for similarity matrix based on correlation to create an edge
Returns
-------
ism : array_like
Individual stability matrix of shape (`V`, `V`), `V` voxels
"""
print(('Calculating individual stability matrix of:', subject_file))
from CPAC.basc import individual_stability_matrix
from CPAC.utils import safe_shape
import nibabel as nb
import numpy as np
import os
data = nb.load(subject_file).get_data().astype('float64')
roi_mask_file = nb.load(roi_mask_file).get_data().astype('float64').astype('bool')
if not safe_shape(roi_mask_file, data):
raise ValueError('Subject %s with volume shape %s conflicts with mask shape %s' % (subject_file,
str(data.shape[:3]),
str(roi_mask_file.shape)) )
Y = data[roi_mask_file].T
print(('(%i timepoints, %i voxels) and %i bootstraps' % (Y.shape[0], Y.shape[1], n_bootstraps)))
ism = individual_stability_matrix(Y, n_bootstraps, k_clusters, cbb_block_size=cbb_block_size, affinity_threshold=affinity_threshold)
ism_file = os.path.join(os.getcwd(), 'individual_stability_matrix.npy')
np.save(ism_file, ism)
print(('Saving individual stability matrix %s for %s' % (ism_file, subject_file)))
return ism_file
def nifti_individual_stability(subject_file, roi_mask_file, n_bootstraps, k_clusters, cbb_block_size = None, affinity_threshold = 0.5):
"""
Calculate the individual stability matrix for a single subject by using Circular Block Bootstrapping method
for time-series data.
Parameters
----------
subject_file : string
Nifti file of a subject
roi_mask_file : string
Region of interest (this method is too computationally intensive to perform on a whole-brain volume)
n_bootstraps : integer
Number of bootstraps
k_clusters : integer
Number of clusters
cbb_block_size : integer, optional
Size of the time-series block when performing circular block bootstrap
affinity_threshold : float, optional
Minimum threshold for similarity matrix based on correlation to create an edge
Returns
-------
ism : array_like
Individual stability matrix of shape (`V`, `V`), `V` voxels
"""
print 'Calculating individual stability matrix of:', subject_file
from CPAC.basc import individual_stability_matrix
from CPAC.utils import safe_shape
import nibabel as nb
import numpy as np
import os
data = nb.load(subject_file).get_data().astype('float64')
roi_mask_file = nb.load(roi_mask_file).get_data().astype('float64').astype('bool')
if not safe_shape(roi_mask_file, data):
raise ValueError('Subject %s with volume shape %s conflicts with mask shape %s' % (subject_file,
str(data.shape[:3]),
str(roi_mask_file.shape)) )
Y = data[roi_mask_file].T
print '(%i timepoints, %i voxels) and %i bootstraps' % (Y.shape[0], Y.shape[1], n_bootstraps)
ism = individual_stability_matrix(Y, n_bootstraps, k_clusters, cbb_block_size=cbb_block_size, affinity_threshold=affinity_threshold)
ism_file = os.path.join(os.getcwd(), 'individual_stability_matrix.npy')
np.save(ism_file, ism)
print 'Saving individual stability matrix %s for %s' % (ism_file, subject_file)
return ism_file
def joint_mask(subjects_file_list, mask_file, dtype='float64'):
"""
Creates a joint mask (intersection) common to all the subjects in a provided list
and a provided mask.
The mask for each subject is created by first computing the standard
deviation at each voxel (i.e., a voxelwise maps of the std across time).
Then this standard deviation map is binarized and used as the mask.
Parameters
----------
subjects_file_list : list of strings
A length `N` list of file paths of the nifti files of subjects
mask_file : string
Path to a mask file in nifti format (this will serve as a prior mask)
dtype : string
Numpy data type, should be one of 'float16', 'float32', or 'float64'
Returns
-------
joint_mask : string
Path to joint mask file in nifti format
(this will be the current directory followed by join_mask.nii.gz)
"""
import nibabel as nb
import numpy as np
import os
from CPAC.utils import safe_shape
mask_img = nb.load(mask_file)
mask = mask_img.get_data().astype('bool')
for subject_file in subjects_file_list:
subj_mask = nb.load(subject_file).get_data().astype(dtype).std(-1)
if not safe_shape(subj_mask, mask): raise ValueError('Subject %s with volume shape %s conflicts \
with mask shape %s' % ( subject_file,
str(subj_mask.shape),
str(mask.shape) ) )
mask *= subj_mask.astype('bool')
print '... joint subject/roi mask loaded'
img = nb.Nifti1Image(mask, header=mask_img.get_header(), affine=mask_img.get_affine())
img_file = os.path.join(os.getcwd(), 'joint_mask.nii.gz')
img.to_filename(img_file)
return img_file
def joint_mask(subjects_file_list, mask_file):
"""
Creates a joint mask (intersection) common to all the subjects in a provided list
and a provided mask
Parameters
----------
subjects_file_list : list of strings
A length `N` list of file paths of the nifti files of subjects
mask_file : string
Path to a mask file in nifti format
Returns
-------
joint_mask : string
Path to joint mask file in nifti format
"""
import nibabel as nb
import numpy as np
import os
from CPAC.utils import safe_shape
nii = nb.load(mask_file)
mask = nii.get_data().astype('bool')
for subject_file in subjects_file_list:
sdata = nb.load(subject_file).get_data().astype(np.float64).sum(-1)
if not safe_shape(sdata, mask):
raise ValueError('Subject %s with volume shape %s conflicts \
with mask shape %s' %
(subject_file, str(sdata.shape), str(mask.shape)))
mask *= sdata.astype('bool')
print '... joint subject/roi mask loaded'
img = nb.Nifti1Image(mask,
header=nii.get_header(),
affine=nii.get_affine())
img_file = os.path.join(os.getcwd(), 'joint_mask.nii.gz')
img.to_filename(img_file)
return img_file
def joint_mask(subjects_file_list, mask_file):
"""
Creates a joint mask (intersection) common to all the subjects in a provided list
and a provided mask
Parameters
----------
subjects_file_list : list of strings
A length `N` list of file paths of the nifti files of subjects
mask_file : string
Path to a mask file in nifti format
Returns
-------
joint_mask : string
Path to joint mask file in nifti format
"""
import nibabel as nb
import numpy as np
import os
from CPAC.utils import safe_shape
nii = nb.load(mask_file)
mask = nii.get_data().astype("bool")
for subject_file in subjects_file_list:
sdata = nb.load(subject_file).get_data().astype(np.float64).sum(-1)
if not safe_shape(sdata, mask):
raise ValueError(
"Subject %s with volume shape %s conflicts \
with mask shape %s"
% (subject_file, str(sdata.shape), str(mask.shape))
)
mask *= sdata.astype("bool")
print "... joint subject/roi mask loaded"
img = nb.Nifti1Image(mask, header=nii.get_header(), affine=nii.get_affine())
img_file = os.path.join(os.getcwd(), "joint_mask.nii.gz")
img.to_filename(img_file)
return img_file
def extract_tissue_data(data_file,
ventricles_mask_file,
wm_seg_file, csf_seg_file, gm_seg_file,
wm_threshold=0.0, csf_threshold=0.0, gm_threshold=0.0):
import numpy as np
import nibabel as nb
import os
from CPAC.nuisance import erode_mask
from CPAC.utils import safe_shape
print('Tissues extraction thresholds wm %d, csf %d, gm %d' % (wm_threshold,
csf_threshold,
gm_threshold))
try:
data = nb.load(data_file).get_data().astype('float64')
except:
raise MemoryError('Unable to load %s' % data_file)
try:
lat_ventricles_mask = nb.load(ventricles_mask_file).get_data().astype('float64')
except:
raise MemoryError('Unable to load %s' % lat_ventricles_mask)
if not safe_shape(data, lat_ventricles_mask):
raise ValueError('Spatial dimensions for data and the lateral ventricles mask do not match')
try:
wm_seg = nb.load(wm_seg_file).get_data().astype('float64')
except:
raise MemoryError('Unable to load %s' % wm_seg)
if not safe_shape(data, wm_seg):
raise ValueError('Spatial dimensions for data, white matter segment do not match')
wm_mask = erode_mask(wm_seg > wm_threshold)
wm_sigs = data[wm_mask]
file_wm = os.path.join(os.getcwd(), 'wm_signals.npy')
np.save(file_wm, wm_sigs)
del wm_sigs
try:
csf_seg = nb.load(csf_seg_file).get_data().astype('float64')
except:
raise MemoryError('Unable to load %s' % csf_seg)
if not safe_shape(data, csf_seg):
raise ValueError('Spatial dimensions for data, cerebral spinal fluid segment do not match')
# Only take the CSF at the lateral ventricles as labeled in the Harvard
# Oxford parcellation regions 4 and 43
csf_mask = (csf_seg > csf_threshold)*(lat_ventricles_mask==1)
csf_sigs = data[csf_mask]
file_csf = os.path.join(os.getcwd(), 'csf_signals.npy')
np.save(file_csf, csf_sigs)
del csf_sigs
try:
gm_seg = nb.load(gm_seg_file).get_data().astype('float64')
except:
raise MemoryError('Unable to load %s' % gm_seg)
if not safe_shape(data, gm_seg):
raise ValueError('Spatial dimensions for data, gray matter segment do not match')
gm_mask = erode_mask(gm_seg > gm_threshold)
gm_sigs = data[gm_mask]
file_gm = os.path.join(os.getcwd(), 'gm_signals.npy')
np.save(file_gm, gm_sigs)
del gm_sigs
nii = nb.load(wm_seg_file)
wm_mask_file = os.path.join(os.getcwd(), 'wm_mask.nii.gz')
csf_mask_file = os.path.join(os.getcwd(), 'csf_mask.nii.gz')
gm_mask_file = os.path.join(os.getcwd(), 'gm_mask.nii.gz')
nb.Nifti1Image(wm_mask, header=nii.get_header(), affine=nii.get_affine()).to_filename(wm_mask_file)
nb.Nifti1Image(csf_mask, header=nii.get_header(), affine=nii.get_affine()).to_filename(csf_mask_file)
nb.Nifti1Image(gm_mask, header=nii.get_header(), affine=nii.get_affine()).to_filename(gm_mask_file)
return file_wm, file_csf, file_gm
def extract_tissue_data(data_file,
ho_mask_file,
wm_seg_file,
csf_seg_file,
gm_seg_file,
wm_threshold=0.0,
csf_threshold=0.0,
gm_threshold=0.0):
import numpy as np
import nibabel as nb
import os
from CPAC.nuisance import erode_mask
from CPAC.utils import safe_shape
print 'Tissues extraction thresholds wm %d, csf %d, gm %d' % (
wm_threshold, csf_threshold, gm_threshold)
try:
data = nb.load(data_file).get_data().astype('float64')
except:
raise MemoryError('Unable to load %s' % data_file)
try:
ho_mask = nb.load(ho_mask_file).get_data().astype('float64')
except:
raise MemoryError('Unable to load %s' % ho_mask)
if not safe_shape(data, ho_mask):
raise ValueError(
'Spatial dimensions for data and ho_mask do not match')
try:
wm_seg = nb.load(wm_seg_file).get_data().astype('float64')
except:
raise MemoryError('Unable to load %s' % wm_seg)
if not safe_shape(data, wm_seg):
raise ValueError(
'Spatial dimensions for data, white matter segment do not match')
wm_mask = erode_mask(wm_seg > wm_threshold)
wm_sigs = data[wm_mask]
file_wm = os.path.join(os.getcwd(), 'wm_signals.npy')
np.save(file_wm, wm_sigs)
del wm_sigs
try:
csf_seg = nb.load(csf_seg_file).get_data().astype('float64')
except:
raise MemoryError('Unable to load %s' % csf_seg)
if not safe_shape(data, csf_seg):
raise ValueError(
'Spatial dimensions for data, cerebral spinal fluid segment do not match'
)
# Only take the CSF at the lateral ventricals as labled in the Harvard
# Oxford parcellation regions 4 and 43
csf_mask = (csf_seg > csf_threshold) * ((ho_mask == 43) + (ho_mask == 4))
csf_sigs = data[csf_mask]
file_csf = os.path.join(os.getcwd(), 'csf_signals.npy')
np.save(file_csf, csf_sigs)
del csf_sigs
try:
gm_seg = nb.load(gm_seg_file).get_data().astype('float64')
except:
raise MemoryError('Unable to load %s' % gm_seg)
if not safe_shape(data, gm_seg):
raise ValueError(
'Spatial dimensions for data, gray matter segment do not match')
gm_mask = erode_mask(gm_seg > gm_threshold)
gm_sigs = data[gm_mask]
file_gm = os.path.join(os.getcwd(), 'gm_signals.npy')
np.save(file_gm, gm_sigs)
del gm_sigs
nii = nb.load(wm_seg_file)
wm_mask_file = os.path.join(os.getcwd(), 'wm_mask.nii.gz')
csf_mask_file = os.path.join(os.getcwd(), 'csf_mask.nii.gz')
gm_mask_file = os.path.join(os.getcwd(), 'gm_mask.nii.gz')
nb.Nifti1Image(wm_mask, header=nii.get_header(),
affine=nii.get_affine()).to_filename(wm_mask_file)
nb.Nifti1Image(csf_mask, header=nii.get_header(),
affine=nii.get_affine()).to_filename(csf_mask_file)
nb.Nifti1Image(gm_mask, header=nii.get_header(),
affine=nii.get_affine()).to_filename(gm_mask_file)
return file_wm, file_csf, file_gm
