import os
import argparse
import numpy as np
import nibabel as nib
from nipy.labs import as_volume_img
from scipy.stats import scoreatpercentile
if __name__ == '__main__':
parser = argparse.ArgumentParser(description='Normalize DWI intensity for group fiber tractography.')
parser.add_argument('-i', dest='dwi_file') # 4D DWI Nifti
dwi_path = parser.parse_args().dwi_file
[output_path, dwi_file] = os.path.split(dwi_path)
dwi_file = dwi_file.replace(".nii", "_norm.nii")
dwi_img = as_volume_img(dwi_path)
dwi = dwi_img.get_data()
dwi = dwi / np.float(scoreatpercentile(dwi.ravel(), 99))
dwi = np.int16(dwi * 1e4)
nii = nib.Nifti1Image(dwi, dwi_img.affine)
nib.save(nii, os.path.join(output_path, dwi_file))
graph_left_ipsi = binary dxd matrix with 1 for spatially-connected parcels on left side of brain (.mat)
"""
import os
import numpy as np
import argparse
from nipy.labs import as_volume_img
from scipy import io
from scipy.ndimage.morphology import binary_dilation
if __name__ == '__main__':
parser = argparse.ArgumentParser(description='Generate adjacency and bilateral graph.')
parser.add_argument('-i', dest='filepath')
filepath = parser.parse_args().filepath
path, afile = os.path.split(filepath)
afile = afile.replace('.nii','.mat')
template_img = as_volume_img(filepath)
template = template_img.get_data()
rois = np.unique(template)
rois = rois[1:] # Remove background
n_rois = np.size(rois)
n_neigh = 1 # Increase n_neigh to increase robustness to parcellation errors
# Build bilateral connection matrix
masks = template[np.newaxis, :] == np.arange(n_rois + 1)[:, np.newaxis, np.newaxis, np.newaxis] # Each row = mask of an ROI
grid = np.mgrid[0:template.shape[0], 0:template.shape[1], 0:template.shape[2]] # Three 3D volumes of x, y, and z coords
roi_cen = (grid[np.newaxis, :] * masks[:, np.newaxis, :]).reshape(n_rois + 1, 3, -1).sum(axis=-1) / masks.reshape(n_rois + 1, -1).sum(axis=-1).astype(np.float64)[:, np.newaxis]
roi_cen = roi_cen[1:] # Compute centroid of each ROI
x_cen = np.mean(grid[0][template > 0]) # x centroid computed at voxel level
roi_cen_xflip = roi_cen.copy()
roi_cen_xflip[:, 0] = 2 * x_cen - roi_cen[:, 0] # Find bilateral ROI
ind_neigh = np.sum((roi_cen[np.newaxis, :] - roi_cen_xflip[:, np.newaxis, :]) ** 2, axis=2).argsort(axis=1)
import os
import numpy as np
import argparse
import nibabel as nib
from nipy.labs import as_volume_img
from scipy.ndimage.morphology import binary_dilation
if __name__ == '__main__':
parser = argparse.ArgumentParser(description='Binarize tissue mask based on user-defined threshold (-t 0.33.')
parser.add_argument('-i', dest='filename') # Take .nii as input
parser.add_argument('-t', dest='threshold', default=0.33)
filepath = parser.parse_args().filename
thresh = np.float32(parser.parse_args().threshold)
path, afile = os.path.split(filepath)
afile = afile.replace('.nii', '_bin.nii')
vol_img = as_volume_img(filepath)
vol = vol_img.get_data() > thresh
vol = binary_dilation(vol, structure=np.ones((2, 2, 2))).astype(np.uint8)
nii = nib.Nifti1Image(vol, vol_img.affine)
nib.save(nii, os.path.join(path, afile))
#for sub in subList:
# tc = io.loadmat(os.path.join(BASE_DIR, sub, "restfMRI/tc_rest_vox.mat"))
# tc = tc["tc_vox"]
# pca = PCA(n_components=10)
# pca.fit(tc.T)
# if sub == subList[0]:
# tc_group = preprocessing.standardize(pca.transform(tc.T))
# else:
# tc_group = np.hstack((tc_group, preprocessing.standardize(pca.transform(tc.T))))
# print("Concatenating subject" + sub + "'s timecourses")
#io.savemat(os.path.join(BASE_DIR, "group/tc_rest_pca_vox.mat"), {"tc_group": tc_group})
tc_group = io.loadmat(os.path.join(BASE_DIR, "group/tc_rest_pca_vox.mat"))
tc_group = tc_group["tc_group"]
# Load Freesurfer template
brain_img = as_volume_img("/volatile/bernardng/templates/freesurfer/cort_subcort_333.nii")
brain = brain_img.get_data()
dim = np.shape(brain)
rois = np.unique(brain)
rois = rois[1:] # Remove background
n_rois = np.shape(rois)[0]
n_vox = np.sum(brain != 0) # Number of voxels within ROIs in Freesurfer template
# Spatially smoothing the group timecourses
tc_group = tc_group.reshape((dim[0], dim[1], dim[2], -1))
n_tpts = tc_group.shape[-1]
for t in np.arange(n_tpts):
tc_group[:,:,:,t] = gaussian_filter(tc_group[:,:,:,t], sigma=5)
tc_group = tc_group.reshape((-1, n_tpts))
# Perform parcellation on smoothed PCA-ed timecourses for each ROI
# Standardizing pca-ed time courses
tc_std = np.std(tc_pca, axis=1)
ind = tc_std > 1e-16
tc_pca[ind, :] = tc_pca[ind, :] - tc_pca[ind, :].mean(axis=1)[:, np.newaxis]
tc_pca[ind, :] = tc_pca[ind, :] / tc_pca[ind, :].std(axis=1)[:, np.newaxis]
# Concatenate time courses across subjects
if sub == subList[0]:
tc_group = tc_pca
else:
tc_group = np.hstack((tc_group, tc_pca))
print ("Concatenating subject" + sub + "'s timecourses")
# Load Freesurfer template
brain_img = as_volume_img(ANAT_DIR)
brain = brain_img.get_data()
dim = np.shape(brain)
rois = np.unique(brain)
rois = rois[1:] # Remove background
n_rois = np.shape(rois)[0]
n_vox = np.sum(brain != 0) # Number of voxels within ROIs in Freesurfer template
# Spatially smoothing the group timecourses
tc_group = tc_group.reshape((dim[0], dim[1], dim[2], -1))
n_tpts = tc_group.shape[-1]
for t in np.arange(n_tpts):
tc_group[:, :, :, t] = gaussian_filter(tc_group[:, :, :, t], sigma=2.5)
tc_group = tc_group.reshape((-1, n_tpts))
# Perform parcellation on smoothed PCA-ed timecourses for each ROI
# Standardizing pca-ed time courses
tc_std = np.std(tc_pca, axis=1)
ind = tc_std > 1e-16
tc_pca[ind, :] = tc_pca[ind, :] - tc_pca[ind, :].mean(axis=1)[:, np.newaxis]
tc_pca[ind, :] = tc_pca[ind, :] / tc_pca[ind, :].std(axis=1)[:, np.newaxis]
# Concatenate time courses across subjects
if sub == subList[0]:
tc_group = tc_pca
else:
tc_group = np.hstack((tc_group, tc_pca))
print("Concatenating subject" + sub + "'s timecourses")
# Load Freesurfer template
brain_img = as_volume_img(ANAT_DIR)
# Run relabel_disconnected_rois.py
brain = brain_img.get_data()
dim = np.shape(brain)
rois = np.unique(brain)
rois = rois[1:] # Remove background
n_rois = np.shape(rois)[0]
n_vox = np.sum(brain != 0) # Number of voxels within ROIs in Freesurfer template
# Spatially smoothing the group timecourses
tc_group = tc_group.reshape((dim[0], dim[1], dim[2], -1))
n_tpts = tc_group.shape[-1]
for t in np.arange(n_tpts):
tc_group[:,:,:,t] = gaussian_filter(tc_group[:,:,:,t], sigma=1.5)
# Add shifted motion regressors
for i in np.array([-1, 1]):
regressors = np.hstack(
(regressors, np.roll(motion_confounds, i, axis=0)))
else:
regressors = io.loadmat(reg_file)
regressors = regressors['SPM'][0, 0].xX[
0, 0].X # Contains task and SHIFTED versions of motion regressors
# # Data from other than IMAGEN database might not have shifted motion regressors
# motion_confounds = regressors[:, n_cond:]
# for i in np.array([-1, 1]):
# regressors = np.hstack((regressors, np.roll(motion_confounds, i, axis=0)))
print "Extracting time series..."
# Extract timeseries
tc_img = as_volume_img(tc_file)
tc = tc_img.get_data()
tc_dim = tc.shape
n_tpts = tc_dim[3]
tc = tc.reshape((-1, n_tpts)).T
tc[np.isnan(tc)] = 0 # Remove NAN's
print "Resampling tissue masks to EPI resolution..."
# Load tissue data objects
gm_img = as_volume_img(gm_file)
gm_img = gm_img.resampled_to_img(tc_img)
gm = gm_img.get_data()
gm[gm < 0] = 0
wm_img = as_volume_img(wm_file)
wm_img = wm_img.resampled_to_img(tc_img)
wm = wm_img.get_data()
s = 1
# Constants
BASE_DIR = "/volatile/bernardng/data/sepideh/preprocessedBN/"
TR = 2.4
subList = np.loadtxt(os.path.join(BASE_DIR, "subListTruncated"), dtype='str')
for sub in [subList[s]]:
# Define path to data
tc_file = os.path.join(BASE_DIR, sub, "rsfMRI", "wbold_audiospont_1.hdr")
gm_file = os.path.join(BASE_DIR, sub, "anat", "wc1anat_" + sub + "_3T_neurospin.hdr")
wm_file = os.path.join(BASE_DIR, sub, "anat", "wc2anat_" + sub + "_3T_neurospin.hdr")
csf_file = os.path.join(BASE_DIR, sub, "anat", "wc3anat_" + sub + "_3T_neurospin.hdr")
motion_confounds = np.loadtxt(os.path.join(BASE_DIR, sub, "rsfMRI", "rp_bold_audiospont_1_0001.txt"))
template_file = "/volatile/bernardng/templates/yeo_sulci_merged/Yeo_cort_sulci_subcort_MNI152_3mm.nii"
# Load data objects
tc_img = as_volume_img(tc_file)
gm_img = as_volume_img(gm_file)
wm_img = as_volume_img(wm_file)
csf_img = as_volume_img(csf_file)
template_img = as_volume_img(template_file)
# Temporal filtering and removing WM and CSF signal
tc, tc_roi = preproc(tc_img, gm_img, wm_img, csf_img, motion_confounds, template_img=template_img, tr=TR)
# Graphical LASSO
# glasso = GraphLassoCV(verbose=1, n_refinements=5, alphas=5, n_jobs=1)
# glasso.fit(tc_roi)
# cov_ = glasso.covariance_
# prec_ = glasso.precision_
Input: template_file = location of parcel template (.nii)
Output: parcel_centroid = centroid of each parcel
"""
import os
import numpy as np
import argparse
from nipy.labs import as_volume_img
from scipy import io
if __name__ == '__main__':
parser = argparse.ArgumentParser(description='Compute parcel centroid.')
parser.add_argument('-i', dest='filepath')
filepath = parser.parse_args().filepath
path, afile = os.path.split(filepath)
afile = afile.replace('.nii','_coords.mat')
template_img = as_volume_img(filepath)
template = template_img.get_data()
rois = np.unique(template)
rois = rois[1:] # Remove background
n_rois = np.size(rois)
masks = template[np.newaxis, :] == np.arange(n_rois + 1)[:, np.newaxis, np.newaxis, np.newaxis] # Each row = mask of an ROI
grid = np.mgrid[0:template.shape[0], 0:template.shape[1], 0:template.shape[2]] # Three 3D volumes of x, y, and z coords
roi_cen = (grid[np.newaxis, :] * masks[:, np.newaxis, :]).reshape(n_rois + 1, 3, -1).sum(axis=-1) / masks.reshape(n_rois + 1, -1).sum(axis=-1).astype(np.float64)[:, np.newaxis]
roi_cen = roi_cen[1:] # Compute centroid of each ROI
roi_cen = np.vstack((roi_cen.T, np.ones(n_rois))
parcel_cen_mni = np.dot(template_img.affine, roi_cen)
parcel_cen_mni = parcel_cen_mni[0:3].T
io.savemat(os.path.join(path, afile), {"coords": parcel_cen_mni})
import os
import numpy as np
from scipy import io
from nipy.labs import as_volume_img
BASE_DIR = "/volatile/bernardng/data/imagen/"
TR = 2.2
#subList = np.loadtxt(os.path.join(BASE_DIR, "subjectLists/subjectList.txt"), dtype='str')
subList = np.loadtxt(os.path.join(BASE_DIR, "subjectLists/facesList.txt"), dtype='str')
data_type = 0 # 0 = task, # 1 = rest
# Load parcel template
template = io.loadmat(os.path.join(BASE_DIR, "group/parcel500refined.mat"))
template = template['template'].ravel()
label = np.unique(template)
brain_img = as_volume_img("/volatile/bernardng/templates/spm8/rgrey.nii") # For resample the tissue maps
for sub in subList:
print str("Subject" + sub)
# Load preprocessed voxel timecourses
if data_type:
tc = io.loadmat(os.path.join(BASE_DIR, sub, "restfMRI/tc_rest_vox.mat"))
else:
# tc = io.loadmat(os.path.join(BASE_DIR, sub, "gcafMRI/tc_task_vox.mat"))
tc = io.loadmat(os.path.join(BASE_DIR, sub, "facesfMRI/tc_task_vox.mat"))
tc = tc["tc_vox"]
# Generate tissue mask
gm_file = os.path.join(BASE_DIR, sub, "anat", "gmMask.nii")
gm_img = as_volume_img(gm_file)
gm_img = gm_img.resampled_to_img(brain_img)
gm = gm_img.get_data()
wm_file = os.path.join(BASE_DIR, sub, "anat", "wmMask.nii")
n_parcels = 500.0
# Change path to files
TC_PATH = "tc_vox.mat"
REF_PATH = "wea000000459848s007a001.nii.gz"
GM_PATH = "../Maps/wc1mprage000000459848.nii.gz"
WM_PATH = "../Maps/wc2mprage000000459848.nii.gz"
CSF_PATH = "../Maps/wc3mprage000000459848.nii.gz"
PARCEL_PATH = "parcel500.nii"
# Load subject timecourse
tc = io.loadmat(TC_PATH)
tc = tc["tc"].T
# Generate dilated GM mask
ref_img = as_volume_img(REF_PATH) # For resampling to 3mm
ref = ref_img.get_data()
gm_img = as_volume_img(GM_PATH)
gm = gm_img.get_data()
#wm_img = as_volume_img(WM_PATH)
#wm = wm_img.get_data()
#csf_img = as_volume_img(CSF_PATH)
#csf = csf_img.get_data()
#probTotal = gm + wm + csf
#dim = np.shape(probTotal)
#ind = probTotal > 0
#gm[ind] = gm[ind] / probTotal[ind]
#wm[ind] = wm[ind] / probTotal[ind]
#csf[ind] = csf[ind] / probTotal[ind]
#tissue = np.array([gm.ravel(), wm.ravel(), csf.ravel()])
#tissue_mask = tissue.argmax(axis=0) + 1
# Standardizing pca-ed time courses
tc_std = np.std(tc_pca, axis=1)
ind = tc_std > 1e-16
tc_pca[ind, :] = tc_pca[ind, :] - tc_pca[ind, :].mean(axis=1)[:, np.newaxis]
tc_pca[ind, :] = tc_pca[ind, :] / tc_pca[ind, :].std(axis=1)[:, np.newaxis]
# Concatenate time courses across subjects
if sub == subList[0]:
tc_group = tc_pca
else:
tc_group = np.hstack((tc_group, tc_pca))
print("Concatenating subject" + sub + "'s timecourses")
# Generate dilated GM mask
ref_img = as_volume_img(REF_DIR) # For resampling to 3mm
ref = ref_img.get_data()
gm_img = as_volume_img(GM_DIR)
gm = gm_img.get_data()
wm_img = as_volume_img(WM_DIR)
wm = wm_img.get_data()
csf_img = as_volume_img(CSF_DIR)
csf = csf_img.get_data()
probTotal = gm + wm + csf
dim = np.shape(probTotal)
ind = probTotal > 0
gm[ind] = gm[ind] / probTotal[ind]
wm[ind] = wm[ind] / probTotal[ind]
csf[ind] = csf[ind] / probTotal[ind]
tissue = np.array([gm.ravel(), wm.ravel(), csf.ravel()])
tissue_mask = tissue.argmax(axis=0) + 1
# Concatenating PCA-ed voxel timecourses across subjects
for sub in subList:
tc = io.loadmat(os.path.join(BASE_DIR, sub, "restfMRI/tc_rest_vox.mat"))
tc = tc["tc_vox"]
pca = PCA(n_components=10)
pca.fit(tc.T)
if sub == subList[0]:
tc_group = preprocessing.standardize(pca.transform(tc.T))
else:
tc_group = np.hstack(
(tc_group, preprocessing.standardize(pca.transform(tc.T))))
print("Concatenating subject" + sub + "'s timecourses")
#io.savemat(os.path.join(BASE_DIR, "group/tc_rest_pca_vox.mat"), {"tc_group": tc_group})
# Perform parcellation on PCA-ed timecourses
brain_img = as_volume_img("/volatile/bernardng/templates/spm8/rgrey.nii")
brain = brain_img.get_data()
dim = np.shape(brain)
brain = brain > 0.2 # Generate brain mask
brain = mask_utils.largest_cc(brain)
mem = Memory(cachedir='.', verbose=1)
# Define connectivity based on brain mask
A = grid_to_graph(n_x=brain.shape[0],
n_y=brain.shape[1],
n_z=brain.shape[2],
mask=brain)
# Create ward object
ward = WardAgglomeration(n_clusters=500, connectivity=A, memory=mem)
tc_group = tc_group.reshape((dim[0], dim[1], dim[2], -1))
n_tpts = tc_group.shape[-1]
for t in np.arange(n_tpts):
"""
Resample input volume to resolution of reference
"""
import os
import nibabel as nib
import numpy as np
import argparse
from nipy.labs import as_volume_img
if __name__ == '__main__':
parser = argparse.ArgumentParser(description='Resample input volume to resolution of reference (-t 0 = continuous interpolation; -t 1 = nearest neighbor')
parser.add_argument('-i', dest='input') # Takes .nii as input
parser.add_argument('-r', dest='reference')
parser.add_argument('-t', dest='interp_type', default=0)
input_path = parser.parse_args().input
ref_path = parser.parse_args().reference
interp_type = np.float32(parser.parse_args().interp_type)
if interp_type == 0:
interp = 'continuous'
elif interp_type == 1:
interp = 'nearest'
path, afile = os.path.split(input_path)
afile = afile.replace('.nii', '_rs.nii')
vol_img = as_volume_img(input_path)
ref_img = as_volume_img(ref_path)
vol_img = vol_img.resampled_to_img(ref_img, interpolation=interp)
vol = vol_img.get_data()
nii = nib.Nifti1Image(vol, ref_img.get_affine())
nib.save(nii, os.path.join(path, afile))
n_parcels = 500.0 # Change path to files TC_PATH = "tc_vox.mat" REF_PATH = "wea000000459848s007a001.nii.gz" GM_PATH = "../Maps/wc1mprage000000459848.nii.gz" WM_PATH = "../Maps/wc2mprage000000459848.nii.gz" CSF_PATH = "../Maps/wc3mprage000000459848.nii.gz" PARCEL_PATH = "parcel500.nii" # Load subject timecourse tc = io.loadmat(TC_PATH) tc = tc["tc"].T # Generate dilated GM mask ref_img = as_volume_img(REF_PATH) # For resampling to 3mm ref = ref_img.get_data() gm_img = as_volume_img(GM_PATH) gm = gm_img.get_data() # wm_img = as_volume_img(WM_PATH) # wm = wm_img.get_data() # csf_img = as_volume_img(CSF_PATH) # csf = csf_img.get_data() # probTotal = gm + wm + csf # dim = np.shape(probTotal) # ind = probTotal > 0 # gm[ind] = gm[ind] / probTotal[ind] # wm[ind] = wm[ind] / probTotal[ind] # csf[ind] = csf[ind] / probTotal[ind] # tissue = np.array([gm.ravel(), wm.ravel(), csf.ravel()]) # tissue_mask = tissue.argmax(axis=0) + 1
import os
import argparse
import numpy as np
import nibabel as nib
from nipy.labs import as_volume_img
if __name__ == '__main__':
parser = argparse.ArgumentParser(description='Put B0 volumes at the top.')
parser.add_argument('-i', dest='dwi_file') # 4D DWI Nifti
parser.add_argument('-b', dest='bval_file') # b-value txt file
parser.add_argument('-d', dest='bvec_file') # gradient table
dwi_path = parser.parse_args().dwi_file
bval_path = parser.parse_args().bval_file
bvec_path = parser.parse_args().bvec_file
dwi_img = as_volume_img(dwi_path)
dwi = dwi_img.get_data()
bval = np.loadtxt(bval_path)
bvec = np.loadtxt(bvec_path)
b0_ave = np.mean(dwi[:, :, :, bval == 0], axis=3)
dwi_reorder = np.concatenate(
(b0_ave[:, :, :, np.newaxis], dwi[:, :, :, bval != 0]), axis=3)
bvec_reorder = np.array([0, 0, 0])
bvec_reorder = np.vstack((bvec_reorder, bvec[bval != 0]))
bval_reorder = np.array([0])
bval_reorder = np.hstack((bval_reorder, bval[bval != 0]))
nii = nib.Nifti1Image(dwi_reorder, dwi_img.affine)
nib.save(nii, dwi_path)
np.savetxt(bvec_path, bvec_reorder, fmt='%1.6f')
np.savetxt(bval_path, bval_reorder, fmt='%1.0f')
"""
Create average volume for generating brain mask and aligning fMRI to dMRI
"""
import os
import nibabel as nib
import numpy as np
import argparse
from nipy.labs import as_volume_img
if __name__ == '__main__':
parser = argparse.ArgumentParser(
description='Compute average of a 4D volume.')
parser.add_argument('-i', dest='filename') # Takes .nii as input
filepath = parser.parse_args().filename
path, afile = os.path.split(filepath)
afile = afile.replace('.nii', '_ave.nii')
vol_img = as_volume_img(filepath)
vol = vol_img.get_data()
mean_vol = np.mean(vol, 3)
nii = nib.Nifti1Image(mean_vol, vol_img.get_affine())
nib.save(nii, os.path.join(path, afile))
parser.add_argument('-tc', dest='tc_file')
parser.add_argument('-gm', dest='gm_file')
parser.add_argument('-wm', dest='wm_file')
parser.add_argument('-csf', dest='csf_file')
parser.add_argument('-t', dest='template_file')
tc_file = parser.parse_args().tc_file
gm_file = parser.parse_args().gm_file
wm_file = parser.parse_args().wm_file
csf_file = parser.parse_args().csf_file
template_file = parser.parse_args().template_file
path, afile = os.path.split(tc_file)
_, template_name = os.path.split(template_file)
template_name = template_name.replace('.nii', '')
# Load parcel template
template_img = as_volume_img(template_file)
template = template_img.get_data().ravel()
label = np.unique(template)
# Load preprocessed voxel timecourses
tc = io.loadmat(tc_file)
tc = tc["tc"]
# Generate tissue mask
gm_img = as_volume_img(gm_file)
gm_img = gm_img.resampled_to_img(template_img)
gm = gm_img.get_data()
gm[gm < 0] = 0
wm_img = as_volume_img(wm_file)
wm_img = wm_img.resampled_to_img(template_img)
wm = wm_img.get_data()
from scipy import stats from scipy import signal from scipy import linalg from nipy.labs import mask as mask_utils from nipy.labs import as_volume_img from nipy.labs import viz BASE_DIR = "/volatile/bernardng/data/sepideh/testSubject" # Path to warped greymatter probabilistic mask gm_file = os.path.join(BASE_DIR, "t1mri", "wc1sga070108233-0012-00001-000160-01.hdr") # Path to randomly chosen EPI volume as reference epi_ref_file = os.path.join(BASE_DIR, "fmri", "swfga070108233-0004-00815-000815-01.hdr") # Paths to all EPI volumes with name matching wildcard epi_files = sorted(glob.glob(os.path.join(BASE_DIR, 'fmri', 'swf*.hdr'))) # Load gm_img and epi_img objects gm_img = as_volume_img(gm_file) epi_ref_img = as_volume_img(epi_ref_file) # Resample tissue mask to grid of epi gm_img = gm_img.resampled_to_img(epi_ref_img) # Extract tissue mask gm = gm_img.get_data() # Normalize the mask to [0,1] gm -= gm.min() gm /= gm.max() # Threshold tissue mask gm_mask = (gm > .5) # Find largest connected component gm_mask = mask_utils.largest_cc(gm_mask) # Extract graymatter voxel timecourses time_series_gm, header_gm = mask_utils.series_from_mask(epi_files, gm_mask)
# Concatenating PCA-ed voxel timecourses across subjects
for sub in subList:
tc = io.loadmat(os.path.join(BASE_DIR, sub, "restfMRI/tc_rest_vox.mat"))
tc = tc["tc_vox"]
pca = PCA(n_components=10)
pca.fit(tc.T)
if sub == subList[0]:
tc_group = preprocessing.standardize(pca.transform(tc.T))
else:
tc_group = np.hstack((tc_group, preprocessing.standardize(pca.transform(tc.T))))
print("Concatenating subject" + sub + "'s timecourses")
#io.savemat(os.path.join(BASE_DIR, "group/tc_rest_pca_vox.mat"), {"tc_group": tc_group})
# Perform parcellation on PCA-ed timecourses
brain_img = as_volume_img("/volatile/bernardng/templates/spm8/rgrey.nii")
brain = brain_img.get_data()
dim = np.shape(brain)
brain = brain > 0.2 # Generate brain mask
brain = mask_utils.largest_cc(brain)
mem = Memory(cachedir='.', verbose=1)
# Define connectivity based on brain mask
A = grid_to_graph(n_x=brain.shape[0], n_y=brain.shape[1], n_z=brain.shape[2], mask=brain)
# Create ward object
ward = WardAgglomeration(n_clusters=500, connectivity=A, memory=mem)
tc_group = tc_group.reshape((dim[0], dim[1], dim[2], -1))
n_tpts = tc_group.shape[-1]
for t in np.arange(n_tpts):
tc_group[:,:,:,t] = gaussian_filter(tc_group[:,:,:,t], sigma=5)
tc_group = tc_group.reshape((-1, n_tpts))
tc_group = tc_group[brain.ravel()==1, :]
import numpy as np
from scipy import io
from nipy.labs import as_volume_img
BASE_DIR = "/volatile/bernardng/data/imagen/"
TR = 2.2
#subList = np.loadtxt(os.path.join(BASE_DIR, "subjectLists/subjectList.txt"), dtype='str')
subList = np.loadtxt(os.path.join(BASE_DIR, "subjectLists/facesList.txt"),
dtype='str')
data_type = 0 # 0 = task, # 1 = rest
# Load parcel template
template = io.loadmat(os.path.join(BASE_DIR, "group/parcel500refined.mat"))
template = template['template'].ravel()
label = np.unique(template)
brain_img = as_volume_img("/volatile/bernardng/templates/spm8/rgrey.nii"
) # For resample the tissue maps
for sub in subList:
print str("Subject" + sub)
# Load preprocessed voxel timecourses
if data_type:
tc = io.loadmat(os.path.join(BASE_DIR, sub,
"restfMRI/tc_rest_vox.mat"))
else:
# tc = io.loadmat(os.path.join(BASE_DIR, sub, "gcafMRI/tc_task_vox.mat"))
tc = io.loadmat(
os.path.join(BASE_DIR, sub, "facesfMRI/tc_task_vox.mat"))
tc = tc["tc_vox"]
# Generate tissue mask
gm_file = os.path.join(BASE_DIR, sub, "anat", "gmMask.nii")
gm_img = as_volume_img(gm_file)
gm_img = gm_img.resampled_to_img(brain_img)