def read_ppms():
"""
Open PPMs (White Matter, Gray Matter, CSF, Rest)
"""
Pdict = {}
for tissue in tissues:
fname = tissue+'PostMap_1000.img'
im = load_image(os.path.join(datadir, fname))
Pdict[tissue] = im.get_data()/1000.
affine = im.get_affine()
return Pdict, affine
def load_probmap(id):
"""
Example:
pm = load_probmap('PostMap_1000.img')
"""
pm = Probmap()
for tissue in tissues:
fname = tissue+id
im = load_image(join(datadir, fname))
tmp = im.get_data().copy()
pm[tissue] = tmp
pm.normalize()
return pm
if len(sys.argv)>1:
similarity = sys.argv[1]
if len(sys.argv)>2:
interp = sys.argv[2]
if len(sys.argv)>3:
optimizer = sys.argv[3]
# Print messages
print ('Source brain: %s' % source)
print ('Target brain: %s' % target)
print ('Similarity measure: %s' % similarity)
print ('Optimizer: %s' % optimizer)
# Get data
print('Fetching image data...')
I = load_image(source_file)
J = load_image(target_file)
# Perform affine registration
# The output is an array-like object such that
# np.asarray(T) is a customary 4x4 matrix
print('Setting up registration...')
tic = time.time()
T = register(I, J,
similarity=similarity,
interp=interp,
optimizer=optimizer)
toc = time.time()
print(' Registration time: %f sec' % (toc-tic))
import numpy as np
from os.path import join
from nipy.io.imageformats import Nifti1Image, load as load_image, save as save_image
from nipy.utils import example_data
from nipy.neurospin.registration import FmriRealign4d
from pylab import *
# Create Nifti1Image instances from both input files
rootpath = 'D:\\home\\Alexis\\data\\eccv'
mask = load_image(join(rootpath, 'mask.nii'))
xyz = np.where(mask.get_data()>0)
run = 'run1'
im_raw = load_image(example_data.get_filename('fiac','fiac0',run+'.nii.gz'))
im_npy = load_image(join(rootpath, 'npy_ra'+run+'.nii.gz'))
im_spm = load_image(join(rootpath, 'ra'+run+'.nii'))
raw = im_raw.get_data()
npy = im_npy.get_data()
spm = im_spm.get_data()
TR = 2.5
def compare(x, y, z):
t = TR*np.arange(raw.shape[3])
plot(t, raw[x,y,z,:], 'r')
plot(t, spm[x,y,z,:], 'g')
plot(t, npy[x,y,z,:], 'b')
from nipy.utils import example_data
from os.path import join, split
import sys
# Optinal argument
iterations = 1
if len(sys.argv)>1:
iterations = int(sys.argv[1])
# Input images are provided with the nipy-data package
runs = ['run1', 'run2']
runnames = [example_data.get_filename('fiac','fiac0',run+'.nii.gz') \
for run in runs]
images = [load_image(runname) for runname in runnames]
# Create Image4d instances -- this is a local class representing a
# series of 3d images
runs = [image4d(im, tr=2.5, slice_order='ascending', interleaved=True) \
for im in images]
# Correct motion within- and between-sessions
# By default 2 within-loops and 5 between-loops will be performed
transforms = realign4d(runs, within_loops=iterations, between_loops=2*iterations)
# Resample data on a regular space+time lattice using 4d interpolation
corr_runs = [resample4d(runs[i], transforms=transforms[i]) for i in range(len(runs))]
# Save images
for i in range(len(runs)):
P = posterior[tissue][posterior.mask()]
Z = P.sum()
tmp = data*P
mu = tmp.sum(1)/Z
mu_ = mu.reshape(2,1)
sigma = np.dot(tmp, data.T)/Z - np.dot(mu_, mu_.T)
gaussians[tissue] = Gaussian(mu, sigma)
return gaussians
# Main Program
# Load prior probability map
print('Loading data...')
prior = load_probmap('_1000Prior.img')
# Load image data and apply masking
im1 = load_image(join(datadir, 'BiasCorIm.img'))
im2 = load_image(join(datadir, 'DistanceMap.img'))
data = np.asarray([im.get_data()[prior.mask()] for im in [im1,im2]])
# Save image data in nifti for visualization with `anatomist`
if save_images:
print('Saving prior image...')
save_image(im1, join(savedir, 'BiasCorIm.nii'))
im = Image(prior.brain_ppm(), affine=im1.get_affine())
save_image(im, join(savedir, 'Prior_Brain.nii'))
# Allocate posterior probability map and initialize it from the prior
posterior = Probmap(prior)
# EM algorithm: refine posterior map
print('Starting EM...')
tmp.remove(',')
cvect = map(float, tmp)
# Input files
fmri_files = [example_data.get_filename('fiac','fiac0',run) for run in ['run1.nii.gz','run2.nii.gz']]
design_files = [example_data.get_filename('fiac','fiac0',run) for run in ['run1_design.npz','run2_design.npz']]
mask_file = example_data.get_filename('fiac','fiac0','mask.nii.gz')
# Get design matrix as numpy array
print('Loading design matrices...')
X = [np.load(f)['X'] for f in design_files]
# Get multi-session fMRI data
print('Loading fmri data...')
Y = [load_image(f) for f in fmri_files]
# Get mask image
print('Loading mask...')
mask = load_image(mask_file)
# GLM fitting
print('Starting fit...')
##glm = LinearModel(Y, X, mask=mask, model='ar1')
glm = LinearModel(Y, X, mask=mask)
# Compute the required contrast
print('Computing test contrast image...')
nregressors = X[0].shape[1] ## should check that all design matrices have the same
c = np.zeros(nregressors)
c[0:4] = cvect
import numpy as np
from os.path import join
from nipy.io.imageformats import load as load_image, save as save_image
from nipy.neurospin.registration import FmriRealign4d
from pylab import *
# Create Nifti1Image instances from both input files
rootpath = 'D:\\home\\Alexis\\data\\karla'
run = 'fms070149316-0032.nii'
raw = load_image(join(rootpath, run))
spm = load_image(join(rootpath, 'spm5_ra'+run))
npy = load_image(join(rootpath, 'new_ra'+run))
raw = raw.get_data()
spm = spm.get_data()
npy = npy.get_data()
def compare(x, y, z):
plot(raw[x,y,z,:], 'g')
plot(spm[x,y,z,:], 'r')
plot(npy[x,y,z,:], 'b')
compare(39, 22, 18)
H[0,1,:] = Hxy[:]
H[1,0,:] = Hxy[:]
Hxz = nd.gaussian_filter(data, sigma=sigma, order=[1,0,1])
H[0,2,:] = Hxz[:]
H[2,0,:] = Hxz[:]
Hyz = nd.gaussian_filter(data, sigma=sigma, order=[0,1,1])
H[1,2,:] = Hyz[:]
H[2,1,:] = Hyz[:]
return H
#datadir = 'D:\home\Alexis\data\delphine\zozo'
datadir = 'D:\Alexis\data\patient_03S0908'
# Read input image
im = load_image(join(datadir, 'BiasCorIm.img'))
data = im.get_data()
data = data.astype('float')
sigma = 3
lda = 1
# Compute hessian
print('Computing Hessian...')
H = hessian(data, sigma=sigma)
# Singular value decomposition
print('Computing singular values...')
S = svd(H)
# Anisotropy measure
print('Computing FA...')
import numpy as np
from os.path import join, split
import sys
import time
from glob import glob
from nipy.io.imageformats import load as load_image, save as save_image
from nipy.neurospin.registration import FmriRealign4d
# Create Nifti1Image instances from both input files
rootpath = "D:\\home\\Alexis\\data\\karla"
runnames = glob(join(rootpath, "fms*.nii"))
runs = [load_image(run) for run in runnames]
## DEBUG
"""
idx = [4,5,6,7]
runs = [runs[i] for i in idx]
runnames = [runnames[i] for i in idx]
"""
print runnames
# Do the job
R = FmriRealign4d(runs, tr=2.4, slice_order="ascending", interleaved=False)
R.correct_motion()
corr_runs = R.resample()
# Save images
for i in range(len(runs)):
aux = split(runnames[i])
save_image(corr_runs[i], join(aux[0], "new_ra" + aux[1]))
from os.path import join
import numpy as np
"""
Example of running affine matching on the 'sulcal2000' database
"""
##rootpath = 'D:\\data\\sulcal2000'
rootpath = '/neurospin/lnao/Panabase/roche/sulcal2000'
print('Scanning data directory...')
# Get data
print('Fetching image data...')
I = load_image(join(rootpath,'nobias_anubis'+'.nii'))
J = load_image(join(rootpath,'ammon_TO_anubis'+'.nii'))
# Setup registration algorithm
matcher = IconicMatcher(I.get_data(), J.get_data(),
I.get_affine(), J.get_affine()) ## I: source, J: target
# Params
size = 5
nsimu = 1
depth = 10
import pylab as pl
# Simulations
for i in range(nsimu):
import numpy as np
import scipy.ndimage as nd
from os.path import join
from nipy.io.imageformats import Nifti1Image as Image, load as load_image, save as save_image
import time
import pylab
id = 5
subjects = ["test", "016", "020", "023_1", "031", "03S0908"]
datadir = join("D:\home\AR203069\data\delphine", subjects[id])
tissues = ["CSF", "GM", "WM"]
# Read input image
im = load_image(join(datadir, "BiasCorIm.img"))
# Compute brain posterior probability map
P = np.zeros(im.get_shape())
for tissue in tissues:
tmp = load_image(join(datadir, tissue + "PostMap_1000.img"))
P += tmp.get_data()
P /= 1000.0
##P = np.gaussian_filter(P, sigma=3)
S = 2 * (P - 0.5)
del P
# Compute gradient norm image
G = nd.gaussian_gradient_magnitude(im.get_data(), sigma=3)
# Speed function
Gn = G / float(G.max())
# Clean brain tissue maps
print('Decrease brain tissue probabilities outside brain...')
brain_prior = np.zeros(ppm['REST'].shape)
crude_brain = 2
## Morphology-based crude brain
if crude_brain == 1:
tmp = closing(ppm['REST'], size=6)
threshold = .5
brain_prior[mask] = np.minimum(1.-tmp[mask]/threshold, 1.)
del tmp
## Template-base crude brain
elif crude_brain in [2,3]:
im = load_image(os.path.join(datadir, 'MaskInitAnnot.img'))
brain_prior = im.get_data()
if crude_brain == 3:
brain_prior = gaussian_filter(brain_prior+.0, size=1)
brain_prior /= brain_prior.max()
## Clean using crude brain mask
ppm = geometry_cleaner(ppm, brain_prior, mask)
ppm, mask = normalize(ppm)
# Smooth using log opinion pool
print('Performing log opinion pool smoothing...')
for tissue in tissues:
ppm[tissue] = log_opinion_pool(ppm[tissue], mask, size=1)
ppm, mask = normalize(ppm)
normalize = None
if len(sys.argv)>5:
normalize = sys.argv[5]
optimizer = 'powell'
if len(sys.argv)>6:
optimizer = sys.argv[6]
# Print messages
print ('Source brain: %s' % source)
print ('Target brain: %s' % target)
print ('Similarity measure: %s' % similarity)
print ('Optimizer: %s' % optimizer)
# Get data
print('Fetching image data...')
I = load_image(join(rootpath,'nobias_'+source+'.nii'))
J = load_image(join(rootpath,'nobias_'+target+'.nii'))
# Perform affine normalization
print('Setting up registration...')
tic = time.time()
T = affine_register(I, J,
similarity=similarity, interp=interp,
normalize=normalize, optimizer=optimizer)
toc = time.time()
print(' Registration time: %f sec' % (toc-tic))
# Resample source image
print('Resampling source image...')
tic = time.time()
