def nilearn_smoothing(in_files, fwhm, out_prefix):
"""
Using nilearn function to do smoothing, it seems better than SPM
Args:
in_files:
fwhm:
out_prefix:
Returns:
"""
from nilearn.image import smooth_img
import os
file_name = os.path.basename(in_files)
if fwhm == 0:
smoothed_data = smooth_img(in_files, None)
else:
smoothed_data = smooth_img(in_files, fwhm)
smoothed_data.to_filename(out_prefix + file_name)
smoothed_files = os.path.abspath(
os.path.join(os.getcwd(), out_prefix + file_name))
return smoothed_files
def transform(self, X, y=None, **kwargs):
if isinstance(X, list) and len(X) == 1:
smoothed_img = smooth_img(X[0], fwhm=self.fwhm)
elif isinstance(X, str):
smoothed_img = smooth_img(X, fwhm=self.fwhm)
else:
smoothed_img = smooth_img(X, fwhm=self.fwhm)
if not self.output_img:
if isinstance(smoothed_img, list):
smoothed_img = np.asarray([img.dataobj for img in smoothed_img])
else:
return smoothed_img.dataobj
return smoothed_img
def create_dataset(self, start_time, end_time, delay, fmri_end, eeg_path,
fmri_path):
vector_exclude = [
'EOG', 'ECG', 'CW1', 'CW2', 'CW3', 'CW4', 'CW5', 'CW6', 'Status'
]
raw = mne.io.read_raw_edf(eeg_path, exclude=vector_exclude)
eeg = raw.get_data()
eeg = mne.filter.filter_data(eeg, sfreq=1000, l_freq=5, h_freq=100)
eeg_flip = np.fliplr(eeg)
fmri_im = image.smooth_img(fmri_path, fwhm=6)
fmri = get_masked_fmri(fmri_im, "sub")
start = start_time
end = start_time + self.segment_length
x_list = []
y_list = []
x_fl_list = []
while end < eeg.shape[1] and end <= fmri_end and end < end_time:
signal = eeg[..., start:end]
signal_flip = eeg_flip[..., start:end]
x = eeg_transform(signal)
y = fmri_transform(end, fmri, delay, fmri_end)
x1 = eeg_transform(signal_flip)
x_list.append(x)
y_list.append(y)
x_fl_list.append(x1)
start += self.step
end += self.step
x_list = np.array(x_list)
x_fl_list = np.array(x_fl_list)
y_list = np.array(y_list)
return x_list, y_list, x_fl_list
def __init__(self, root_dir, file_path, list_IDs, labels, format, transform=None):
# event_file could be added when real data is available.
# add this argument later: tsv_file
"""
Args:
file_path (string): Path to the (fmri) file (nifti or npy).
root_dir (string): Directory with all the images.
transform (callable, optional): Optional transform to be applied
on a sample, i.e. flattening and subsampling.
"""
self.root_dir = root_dir
self.file_path = file_path
self.labels = labels
self.path = os.path.join(self.root_dir, self.file_path)
if format is 'nifti':
self.subject_frames = smooth_img(self.path, fwhm=None).get_fdata()
self.subject_frames = self.subject_frames[:, :, :, list_IDs]
else:
images = np.load(self.path, encoding='bytes')
images = images[list_IDs, :, :, :]
images = np.reshape(images, (images.shape[1], images.shape[2], images.shape[3], images.shape[0]))
self.subject_frames = images
self.transform = transform
def test_multi_subject_smoothing(self):
# nilearn
from nilearn.image import smooth_img
nilearn_smoothed_img = smooth_img(self.X[0:3], fwhm=[3, 3, 3])
nilearn_smoothed_array = nilearn_smoothed_img[1].dataobj
# photon
smoother = PipelineElement("SmoothImages", hyperparameters={}, fwhm=3)
photon_smoothed_array, _, _ = smoother.transform(self.X[0:3])
branch = NeuroBranch("NeuroBranch", output_img=True)
branch += smoother
photon_smoothed_img, _, _ = branch.transform(self.X[0:3])
# assert
self.assertIsInstance(photon_smoothed_array, np.ndarray)
self.assertIsInstance(photon_smoothed_img[0], Nifti1Image)
self.assertTrue(
np.array_equal(photon_smoothed_array[1], nilearn_smoothed_array)
)
self.assertTrue(
np.array_equal(
photon_smoothed_img[1].dataobj, nilearn_smoothed_img[1].dataobj
)
)
def p_map(task, run, p_values_3d, threshold=0.05):
"""
Generate three thresholded p-value maps.
Parameters
----------
task: int
Task number
run: int
Run number
p_value_3d: 3D array of p_value.
threshold: The cutoff value to determine significant voxels.
Returns
-------
threshold p-value images
"""
fmri_img = image.smooth_img('../../../data/sub001/BOLD/' + 'task00' +
str(task) + '_run00' + str(run) +
'/filtered_func_data_mni.nii.gz',
fwhm=6)
mean_img = image.mean_img(fmri_img)
log_p_values = -np.log10(p_values_3d)
log_p_values[np.isnan(log_p_values)] = 0.
log_p_values[log_p_values > 10.] = 10.
log_p_values[log_p_values < -np.log10(threshold)] = 0
plot_stat_map(nib.Nifti1Image(log_p_values, fmri_img.get_affine()),
mean_img,
title="Thresholded p-values",
annotate=False,
colorbar=True)
def smoothImages(self, images, threshold):
smoothedImages = []
for i in range(len(images)):
smoothedImage = smooth_img(images[i], threshold)
smoothedImages.append(smoothedImage)
print("Image Smoothened")
return smoothedImages
def preprocess(num, subj, subj_dir, subj_warp_dir, force_warp=False):
bold_path = 'BOLD/task001_run00%i/bold_dico_bold7Tp1_to_subjbold7Tp1.nii.gz' % (num+1)
bold_path = os.path.join(DATA_DIR, subj, bold_path)
template_path = os.path.join(DATA_DIR, 'templates', 'grpbold7Tp1', 'brain.nii.gz')
warp_path = os.path.join(DATA_DIR, subj, 'templates', 'bold7Tp1', 'in_grpbold7Tp1', 'subj2tmpl_warp.nii.gz')
output_path = os.path.join(subj_warp_dir, 'run00%i.nii.gz' % num)
if force_warp or not os.path.exists(output_path):
print 'Warping image #%i...' % num
subprocess.call(['fsl5.0-applywarp', '-i', bold_path, '-o', output_path, '-r', template_path, '-w', warp_path, '-d', 'float'])
else:
print 'Reusing cached warp image #%i' % num
print 'Loading image #%i...' % num
bold = load(output_path)
masker = NiftiMasker(load(MASK_FILE))
# masker = niftimasker(load(MASK_FILE), detrend=true, smoothing_fwhm=4.0,
# high_pass=0.01, t_r=2.0, standardize=true)
masker.fit()
print 'Removing confounds from image #%i...' % num
data = masker.transform(bold, confounds(num, subj))
print 'Detrending image #%i...' % num
filtered = np.float32(savgol_filter(data, 61, 5, axis=0))
img = masker.inverse_transform(data-filtered)
print 'Smoothing image #%i...' % num
img = image.smooth_img(img, 4.0)
print 'Saving image #%i...' % num
save(img, os.path.join(subj_dir, 'run00%i.nii.gz' % num))
print 'Finished with image #%i' % num
def preprocess(self, imgs):
smooth_prefix = '' if self.smoothing_fwhm is None else 's%g' % self.smoothing_fwhm
resample_prefix = '' if self.resampling is None else 'r%g' % self.resampling
if not isinstance(imgs, list):
imgs = [imgs]
path_first = imgs[0] if isinstance(imgs[0], str) else imgs[0].get_filename()
path_first_resampled = os.path.join(os.path.dirname(path_first), resample_prefix + os.path.basename(path_first))
path_first_smoothed = os.path.join(os.path.dirname(path_first), smooth_prefix + resample_prefix + os.path.basename(path_first))
if self.resampling is not None or self.smoothing_fwhm is not None:
if self.resampling is not None and not os.path.exists(path_first_smoothed):
if not os.path.exists(path_first_resampled):
imgs = resample_img(imgs, target_affine=np.diag(self.resampling * np.ones(3)))
else:
imgs = [os.path.join(os.path.dirname(img), resample_prefix + os.path.basename(img)) if isinstance(img, str)
else os.path.join(os.path.dirname(img.get_filename()), resample_prefix + os.path.basename(img.get_filename())) for img in imgs]
if self.smoothing_fwhm is not None:
if not os.path.exists(path_first_smoothed):
imgs = smooth_img(imgs, self.smoothing_fwhm)
else:
imgs = [os.path.join(os.path.dirname(img), smooth_prefix + resample_prefix + os.path.basename(img)) if isinstance(img, str)
else os.path.join(os.path.dirname(img.get_filename()), smooth_prefix + resample_prefix + os.path.basename(img.get_filename())) for img in imgs]
else:
imgs = [check_niimg_3d(img) for img in imgs]
return imgs
def transform(self, imgs, confounds=None):
"""
Parameters
----------
imgs: list of Niimg-like objects
"""
self._check_fitted()
if self.smoothing_fwhm:
imgs = smooth_img(imgs, self.smoothing_fwhm)
imgs = [_utils.check_niimg_3d(img) for img in imgs]
for i, roi in enumerate(self.mask_img_):
masker = NiftiMasker(mask_img=roi)
x = masker.fit_transform(imgs)
if self.extract_funcs is not None:
x = np.array([FDICT[f][0](x, **FDICT[f][1]) for f in self.extract_funcs])
if i == 0:
X = x
else:
X = np.concatenate((X, x), axis=0)
return X.swapaxes(0, 1)
def load_fmri(filename, maskname='', sigma=3):
"""
Reads 4D fmri data.
Smooths using 3D Gaussian filter
Applies mask to data:
- If mask not provided, calculates binary mask
returns fmri data matrix (Time x Voxels)
"""
img = nib.load(filename)
print(img.shape)
rep_time = img.header.get_zooms()[-1]
img = image.smooth_img(img, sigma)
if maskname != '':
img_mask = nib.load(maskname)
else:
print('Mask not provided. Calculating mask ...')
img_mask = masking.compute_background_mask(img)
img = masking.apply_mask(img, img_mask)
print('Mask applied!')
print('Detrending data!')
img = signal.clean(img,
detrend=True,
high_pass=0.01,
standardize=False,
t_r=rep_time)
return img
def transformation(filename):
from nilearn import plotting
from nilearn import datasets
import scipy.ndimage as ndimage
img = nib.load(filename)
print(img.get_fdata().shape)
flip_x = np.flip(img.get_fdata(), 0)
flip_y = np.flip(img.get_fdata(), 1)
flip_z = np.flip(img.get_fdata(), 2)
flip_img_x = nib.Nifti1Image(flip_x, affine=np.eye(4))
flip_img_y = nib.Nifti1Image(flip_y, affine=np.eye(4))
flip_img_z = nib.Nifti1Image(flip_z, affine=np.eye(4))
gaussian_img = nib.Nifti1Image(ndimage.gaussian_filter(img.get_fdata(),
sigma=(5, 5, 0),
order=0),
affine=np.eye(4))
from nilearn import image
smooth_img = image.smooth_img(img, [50.1, 10.0, 0.2])
nib.save(flip_img_x,
os.path.join('images', 'flip_x.' + filename.split('/')[-1]))
nib.save(flip_img_y,
os.path.join('images', 'flip_y.' + filename.split('/')[-1]))
nib.save(flip_img_z,
os.path.join('images', 'flip_z.' + filename.split('/')[-1]))
nib.save(smooth_img,
os.path.join('images', 'smooth.' + filename.split('/')[-1]))
print(type(smooth_img))
print(type(gaussian_img))
nib.save(gaussian_img,
os.path.join('images', 'gaussian.' + filename.split('/')[-1]))
def transform(self, imgs, confounds=None):
smooth_prefix = '' if self.smoothing_fwhm is None else 's%g' % self.smoothing_fwhm
resample_prefix = '' if self.smoothing_fwhm is None else 'r%g' % self.smoothing_fwhm
if not isinstance(imgs, list):
imgs = [imgs]
path_first = imgs[0] if isinstance(imgs[0], str) else imgs[0].get_filename()
path_first_resampled = os.path.join(os.path.dirname(path_first), resample_prefix + os.path.basename(path_first))
path_first_smoothed = os.path.join(os.path.dirname(path_first), smooth_prefix + resample_prefix + os.path.basename(path_first))
if self.resampling is not None and self.smoothing_fwhm is not None:
if self.resampling is not None:
if not os.path.exists(path_first_resampled) and not os.path.exists(path_first_smoothed):
imgs = resample_img(imgs, target_affine=np.diag(self.resampling * np.ones(3)))
else:
imgs = []
if self.smoothing_fwhm is not None:
if not os.path.exists(path_first_smoothed):
imgs = smooth_img(imgs, self.smoothing_fwhm)
else:
imgs = []
else:
imgs = [check_niimg_3d(img) for img in imgs]
return self.masker.transform(imgs)
def random_blur(image, mean, std):
"""
mean: mean fwhm in millimeters.
std: standard deviation of fwhm in millimeters.
"""
return smooth_img(image,
fwhm=np.abs(np.random.normal(mean, std, 3)).tolist())
def __main__():
volume = image.index_img("E:\\Users\\Niall\\Documents\\Computer Science\\FinalYearProject\\data\\ds105_raw\\ds105\\sub001\\BOLD\\task001_run001\\bold.nii.gz", 0)
smoothed_img = image.smooth_img(volume, fwhm=5)
# print("Read the images");
plotting.plot_glass_brain(volume, title='plot_glass_brain',
black_bg=True, display_mode='xz')
plotting.plot_glass_brain(volume, title='plot_glass_brain',
black_bg=False, display_mode='xz')
plt.show()
# print("Finished");
#gemerate some numbers
t = np.linspace(1, 10, 2000) # 2000 points between 1 and 10
t
#plot the graph
plt.plot(t, np.cos(t))
plt.ylabel('Subject Response')
plt.show()
def smooth(img):
# we need to preserve the original header because the smoothing function
# f***s the TR up
nimg = smooth_img(img, fwhm=2.0)
return nb.Nifti1Image(nimg.get_data(),
img.get_affine(),
header=img.get_header())
def test_002(self):
from roistats import collect
from nilearn import plotting as nip
from nilearn import image
from nilearn import datasets
from roistats import atlases, plotting
import random
import numpy as np
import pandas as pd
atlas = datasets.load_mni152_brain_mask()
atlas_fp = datasets.fetch_atlas_harvard_oxford('cort-maxprob-thr50-2mm')['maps']
try:
labels = atlases.labels('HarvardOxford-Cortical.xml')
except Exception:
from nilearn import datasets
name = 'cort-maxprob-thr50-2mm'
labels = datasets.fetch_atlas_harvard_oxford(name)['labels']
# Store them in a dict
labels = dict(enumerate(labels))
print(atlas_fp)
images = []
for i in range(0, 50):
d = np.random.rand(*atlas.dataobj.shape) * atlas.dataobj
im = image.new_img_like(atlas, d)
im = image.smooth_img(im, 5)
_, f = tempfile.mkstemp(suffix='.nii.gz')
im.to_filename(f)
images.append(f)
for each in images[:3]:
nip.plot_roi(atlas_fp, bg_img=each)
df = collect.roistats_from_maps(images, atlas_fp, subjects=None)
df = df.rename(columns=labels)
df['group'] = df.index
df['age'] = df.apply(lambda row: random.random()*5+50, axis=1)
df['group'] = df.apply(lambda row: row['group'] < len(df)/2, axis=1)
df['index'] = df.index
plotting.boxplot('Temporal Pole', df, covariates=['age'], by='group')
_ = plotting.lmplot('Temporal Pole', 'age', df, covariates=[],
hue='group', palette='default')
cov = df[['age', 'group']]
melt = pd.melt(df, id_vars='index', value_vars=labels.values(),
var_name='region').set_index('index')
melt = melt.join(cov)
plotting.hist(melt, by='group', region_colname='region',
covariates=['age'])
print('Removing images')
import os
for e in images:
os.unlink(e)
def smooth(img):
# we need to preserve the original header because the smoothing function
# f***s the TR up
nimg = smooth_img(img, fwhm=2.0)
return nb.Nifti1Image(nimg.get_data(),
img.get_affine(),
header=img.get_header())
def test_single_subject_smoothing(self):
# nilearn
nilearn_smoothed_img = smooth_img(self.X[0], fwhm=[3, 3, 3])
nilearn_smoothed_array = nilearn_smoothed_img.dataobj
# photon
smoother = PipelineElement('SmoothImages',
hyperparameters={},
fwhm=3,
batch_size=1)
photon_smoothed_array, _, _ = smoother.transform(self.X[0])
branch = NeuroBranch('NeuroBranch', output_img=True)
branch += smoother
photon_smoothed_img, _, _ = branch.transform(self.X[0])
# assert
self.assertIsInstance(photon_smoothed_array, np.ndarray)
self.assertIsInstance(photon_smoothed_img, Nifti1Image)
self.assertTrue(
np.array_equal(photon_smoothed_array, nilearn_smoothed_array))
self.assertTrue(
np.array_equal(photon_smoothed_img.dataobj,
nilearn_smoothed_img.dataobj))
def p_map(task, run, p_values_3d, threshold=0.05):
"""
Generate three thresholded p-value maps.
Parameters
----------
task: int
Task number
run: int
Run number
p_value_3d: 3D array of p_value.
threshold: The cutoff value to determine significant voxels.
Returns
-------
threshold p-value images
"""
fmri_img = image.smooth_img('../../../data/sub001/BOLD/' + 'task00' +
str(task) + '_run00' + str(run) +
'/filtered_func_data_mni.nii.gz',
fwhm=6)
mean_img = image.mean_img(fmri_img)
log_p_values = -np.log10(p_values_3d)
log_p_values[np.isnan(log_p_values)] = 0.
log_p_values[log_p_values > 10.] = 10.
log_p_values[log_p_values < -np.log10(threshold)] = 0
plot_stat_map(nib.Nifti1Image(log_p_values, fmri_img.get_affine()),
mean_img, title="Thresholded p-values",
annotate=False, colorbar=True)
def transform(self, imgs, confounds=None):
"""
Parameters
----------
imgs: list of Niimg-like objects
"""
self._check_fitted()
if self.smoothing_fwhm:
imgs = smooth_img(imgs, self.smoothing_fwhm)
imgs = [_utils.check_niimg_3d(img) for img in imgs]
for i, roi in enumerate(self.mask_img_):
masker = NiftiMasker(mask_img=roi)
x = masker.fit_transform(imgs)
if self.extract_funcs is not None:
x = np.array([
FDICT[f][0](x, **FDICT[f][1]) for f in self.extract_funcs
])
if i == 0:
X = x
else:
X = np.concatenate((X, x), axis=0)
return X.swapaxes(0, 1)
def gaussian_coord_smoothing(coords,
mask_img=None,
target_affine=None,
fwhm=9.0):
masker = get_masker(mask_img, target_affine)
peaks_img = coords_to_peaks_img(coords, mask_img=masker.mask_img_)
img = image.smooth_img(peaks_img, fwhm=fwhm)
return masker.inverse_transform(masker.transform(img).squeeze())
def get_avmovie_data(sub, run):
"""Get a clean nifti image for one subject and one avmovie run"""
src = NIFTI_SRC.format(sub=sub, run=run)
print('Reading {}'.format(src))
nft = image.smooth_img(image.clean_img(src), data.SMOOTHING)
print('Done')
return nft
def smooth_img(in_file, fwhm, out_file=None):
""" Use nilearn.image.smooth_img.
Returns
-------
out_file: str
The absolute path to the output file.
"""
import nilearn.image as niimg
return niimg.smooth_img(in_file, fwhm=fwhm)
def smooth_img(in_file, fwhm, out_file=None):
""" Use nilearn.image.smooth_img.
Returns
-------
out_file: str
The absolute path to the output file.
"""
import nilearn.image as niimg
return niimg.smooth_img(in_file, fwhm=fwhm)
def _run_interface(self, runtime):
from nilearn import image
import nibabel as nib
import numpy as np
smoothed_img = image.smooth_img(self.inputs.in_file, self.inputs.fwhm)
nib.save(nib.Nifti1Image(smoothed_img, nib.load(self.inputs.in_file).affine), 'smoothed.nii.gz')
return runtime
def smoothNi(PATH_GZ, fwhm):
from nilearn import image
import os
F_smooth = image.smooth_img(PATH_GZ, fwhm=fwhm)
OutFile = 's' + PATH_GZ[PATH_GZ.rfind('/') + 1:]
F_smooth.to_filename(OutFile)
out_file = os.path.abspath(OutFile)
return out_file
def create_smoothed_image(self,
fwhm=4,
in_prefix='mri/mwp1{}.nii',
out_prefix='mri_smoothed/{}.nii'):
for person in self.persons:
path = os.path.join(self.file_dir,
in_prefix.format(person.filename))
origin_nii = nib.load(path)
nii = image.smooth_img(origin_nii, fwhm)
self.save_nii(person, nii, nii_prefix=out_prefix)
def transform(self, imgs, confounds=None):
if self.smoothing_fwhm is not None or self.target_affine is not None:
if self.smoothing_fwhm is not None:
imgs = smooth_img(imgs, self.smoothing_fwhm)
if self.target_affine is not None:
imgs = resample_img(imgs, target_affine=self.target_affine)
else:
imgs = [check_niimg_3d(img) for img in imgs] if isinstance(imgs, list) else check_niimg_3d(imgs)
return imgs
def dump_comps(masker,
compressor,
components,
threshold=2,
fwhm=None,
perc=None):
from scipy.stats import zscore
from nilearn.plotting import plot_stat_map
from nilearn.image import smooth_img
from scipy.stats import scoreatpercentile
if isinstance(compressor, basestring):
comp_name = compressor
else:
comp_name = compressor.__str__().split('(')[0]
for i_c, comp in enumerate(components):
path_mask = op.join(WRITE_DIR,
'%s_%i-%i' % (comp_name, n_comp, i_c + 1))
nii_raw = masker.inverse_transform(comp)
nii_raw.to_filename(path_mask + '.nii.gz')
comp_z = zscore(comp)
if perc is not None:
cur_thresh = scoreatpercentile(np.abs(comp_z), per=perc)
path_mask += '_perc%i' % perc
print('Applying percentile %.2f (threshold: %.2f)' %
(perc, cur_thresh))
else:
cur_thresh = threshold
path_mask += '_thr%.2f' % cur_thresh
print('Applying threshold: %.2f' % cur_thresh)
nii_z = masker.inverse_transform(comp_z)
gz_path = path_mask + '_zmap.nii.gz'
nii_z.to_filename(gz_path)
plot_stat_map(gz_path,
bg_img='colin.nii',
threshold=cur_thresh,
cut_coords=(0, -2, 0),
draw_cross=False,
output_file=path_mask + 'zmap.png')
# optional: do smoothing
if fwhm is not None:
nii_z_fwhm = smooth_img(nii_z, fwhm=fwhm)
gz_mm_path = path_mask + '_zmap_%imm.nii.gz' % fwhm
nii_z_fwhm.to_filename(gz_mm_path)
plot_stat_map(nii_z_fwhm,
bg_img='colin.nii',
threshold=cur_thresh,
cut_coords=(0, -2, 0),
draw_cross=False,
output_file=path_mask + ('zmap_%imm.png' % fwhm))
def transform(self, imgs, confounds=None):
if self.smoothing_fwhm is not None or self.target_affine is not None:
if self.smoothing_fwhm is not None:
imgs = smooth_img(imgs, self.smoothing_fwhm)
if self.target_affine is not None:
imgs = resample_img(imgs, target_affine=self.target_affine)
else:
imgs = [check_niimg_3d(img) for img in imgs] if isinstance(
imgs, list) else check_niimg_3d(imgs)
return imgs
def subject_data(sub):
sessions = np.zeros(360)
sessions[:90] = 1
sessions[90:180] = 2
sessions[180:270] = 3
sessions[270:] = 4
return image.smooth_img(
image.clean_img(image.concat_imgs(
src.format(sub=sub) for src in NIFTI_SRC),
sessions=sessions), SMOOTHING)
def smooth_concat_files(concat_files, fwhm=4.4, verbose=False, rerun=True):
filenames = []
for filey in concat_files:
if verbose: print("Smoothing %s" % k)
smoothed = image.smooth_img(filey, fwhm)
smooth_name = filey.replace('concat',
'concatsmoothed-fwhm%s' % str(fwhm))
if rerun or not (path.exists(smooth_name)):
smoothed.to_filename(smooth_name)
filenames.append(smooth_name)
return filenames
def make_roi_mask(atlas,
betas_example,
roi_id,
fwhm=None,
interpolation='nearest'):
'''
Extract ROI-specific mask from Atlas in T1w space and sample it down to have
the dimensions of the to-be-masked image of beta coefficients
Args:
atlas (Nifti1Image):
Atlas with voxel values indicative of ROI identity
betas_example (Nifti1Image):
Example image of GLM coefficients which will be masked later (only needed for shape and affine)
roi_id (int):
ROI-specific integer used in "atlas"
fwhm (float OR np.ndarray OR None):
FWHM in mm (along each axis, axis-specific OR skip smoothing)
interpolation (str):
Interpolation method by nilearn.image.resample_img (consult nilearn documentation for other options)
Returns:
mask_nifti_r (Nifti1Image):
resampled ROI-specific mask image (target dimensions)
mask_nifti_s (Nifti1Image):
smoothed ROI-specific mask image (original dimensions)
mask_nifti (Nifti1Image):
original binary ROI-specific mask image
'''
# Extract atlas and target nii data
atlas_array = atlas.get_fdata()
atlas_affine = atlas.affine.copy()
betas_array = betas_example.get_fdata()
betas_affine = betas_example.affine.copy()
# Extract ROI-specific mask and resample it
mask = (atlas_array == roi_id).astype(float)
mask_nifti = nifti1.Nifti1Image(mask, atlas_affine)
# Gaussian smoothing of mask
if fwhm is not None:
mask_nifti_s = image.smooth_img(mask_nifti, fwhm)
else:
mask_nifti_s = mask_nifti
# Resample mask
mask_nifti_r = image.resample_img(mask_nifti_s,
target_affine=betas_affine,
target_shape=betas_array.shape,
interpolation=interpolation,
copy=True)
return mask_nifti_r, mask_nifti_s, mask_nifti
def preprocess(self, imgs):
smooth_prefix = '' if self.smoothing_fwhm is None else 's%g' % self.smoothing_fwhm
resample_prefix = '' if self.resampling is None else 'r%g' % self.resampling
if not isinstance(imgs, list):
imgs = [imgs]
path_first = imgs[0] if isinstance(imgs[0],
str) else imgs[0].get_filename()
path_first_resampled = os.path.join(
os.path.dirname(path_first),
resample_prefix + os.path.basename(path_first))
path_first_smoothed = os.path.join(
os.path.dirname(path_first),
smooth_prefix + resample_prefix + os.path.basename(path_first))
if self.resampling is not None or self.smoothing_fwhm is not None:
if self.resampling is not None and not os.path.exists(
path_first_smoothed):
if not os.path.exists(path_first_resampled):
imgs = resample_img(imgs,
target_affine=np.diag(self.resampling *
np.ones(3)))
else:
imgs = [
os.path.join(os.path.dirname(img), resample_prefix +
os.path.basename(img))
if isinstance(img, str) else os.path.join(
os.path.dirname(
img.get_filename()), resample_prefix +
os.path.basename(img.get_filename()))
for img in imgs
]
if self.smoothing_fwhm is not None:
if not os.path.exists(path_first_smoothed):
imgs = smooth_img(imgs, self.smoothing_fwhm)
else:
imgs = [
os.path.join(
os.path.dirname(img), smooth_prefix +
resample_prefix + os.path.basename(img))
if isinstance(img, str) else os.path.join(
os.path.dirname(img.get_filename()),
smooth_prefix + resample_prefix +
os.path.basename(img.get_filename()))
for img in imgs
]
else:
imgs = [check_niimg_3d(img) for img in imgs]
return imgs
def make_roi_mask(seed, ref_affine, fwhm=5):
""" """
if len(seed) == 3:
seed = np.hstack((seed, 1))
pos = np.dot(np.linalg.inv(ref_affine), seed).astype(np.int)[:3]
seed_mask = np.zeros(ref_shape)
seed_mask[pos[0], pos[1], pos[2]] = 1.
seed_img = nib.Nifti1Image(seed_mask, ref_affine)
seed_img = smooth_img(seed_img, fwhm)
seed_img = math_img('img > 1.e-6', img=seed_img)
return seed_img
def nii2jpg(inFile=None,
outFile=None,
cutCoords=(3, 3, 3),
displayMode='ortho'):
epiImage = image.mean_img(inFile)
epiImage = image.smooth_img(epiImage, 'fast')
plotting.plot_epi(epi_img=epiImage,
cut_coords=cutCoords,
output_file=outFile,
display_mode=displayMode,
annotate=False,
draw_cross=False)
def compute_mean_epi(fmri_filenames, output_pathway):
fmri_img = smooth_img(fmri_filenames, fwhm=5)
mean_epi = mean_img(fmri_img)
# Save
mean_epi.to_filename(os.path.join(output_pathway, 'mean_epi.nii.gz'))
# Plot
plotting.plot_epi(mean_epi, title='Smoothed mean EPI',
output_file=os.path.join(output_pathway, 'mean_epi.png'))
return mean_epi
def coords_to_img(coords, fwhm=9.0):
mask = load_mni152_brain_mask()
masker = NiftiMasker(mask).fit()
voxels = np.asarray(
image.coord_transform(*coords.T, np.linalg.pinv(mask.affine)),
dtype=int,
).T
peaks = np.zeros(mask.shape)
np.add.at(peaks, tuple(voxels.T), 1.0)
peaks_img = image.new_img_like(mask, peaks)
img = image.smooth_img(peaks_img, fwhm=fwhm)
img = masker.inverse_transform(masker.transform(img).squeeze())
return img
def save_fake_map(orig_inside_mask, map_nii, row, mask_nii, t_quantiles, t_values, i, j):
np.random.seed(j)
new_data = np.random.choice(orig_inside_mask, size=map_nii.shape, replace=True)
shuffled_nii = nb.Nifti1Image(new_data, map_nii.affine, map_nii.header)
smoothed_nii = smooth_img(shuffled_nii, np.array([row.FWHMx_mm, row.FWHMy_mm, row.FWHMz_mm]))
new_data = smoothed_nii.get_data()
new_data[mask_nii.get_data() != 1] = np.nan
new_inside_mask = stats.zscore(new_data[mask_nii.get_data() == 1])
new_data[mask_nii.get_data() == 1] = new_inside_mask #hist_match(new_inside_mask, t_quantiles, t_values)
masked_nii = nb.Nifti1Image(new_data, map_nii.affine, map_nii.header)
masked_nii.to_filename(data_location +"/images/fake_maps/%04d/%04d.nii.gz"%(i, j))
def dump_comps(masker, compressor, components, threshold=2, fwhm=None,
perc=None):
from scipy.stats import zscore
from nilearn.plotting import plot_stat_map
from nilearn.image import smooth_img
from scipy.stats import scoreatpercentile
if isinstance(compressor, basestring):
comp_name = compressor
else:
comp_name = compressor.__str__().split('(')[0]
for i_c, comp in enumerate(components):
path_mask = op.join(WRITE_DIR, '%s_%i-%i' % (comp_name,
n_comp, i_c + 1))
nii_raw = masker.inverse_transform(comp)
nii_raw.to_filename(path_mask + '.nii.gz')
comp_z = zscore(comp)
if perc is not None:
cur_thresh = scoreatpercentile(np.abs(comp_z), per=perc)
path_mask += '_perc%i' % perc
print('Applying percentile %.2f (threshold: %.2f)' % (perc, cur_thresh))
else:
cur_thresh = threshold
path_mask += '_thr%.2f' % cur_thresh
print('Applying threshold: %.2f' % cur_thresh)
nii_z = masker.inverse_transform(comp_z)
gz_path = path_mask + '_zmap.nii.gz'
nii_z.to_filename(gz_path)
plot_stat_map(gz_path, bg_img='colin.nii', threshold=cur_thresh,
cut_coords=(0, -2, 0), draw_cross=False,
output_file=path_mask + 'zmap.png')
# optional: do smoothing
if fwhm is not None:
nii_z_fwhm = smooth_img(nii_z, fwhm=fwhm)
plot_stat_map(nii_z_fwhm, bg_img='colin.nii', threshold=cur_thresh,
cut_coords=(0, -2, 0), draw_cross=False,
output_file=path_mask +
('zmap_%imm.png' % fwhm))
def make_new_noise(masker):
# CONST
n_random_niis = 1000
n_random_foci = 200, 300 # random amount of foci per study
fwhm_range = (4, 20)
print "Inventing NOISE images from Gaussians..."
# generate some noise niftis
inds = np.where(masker.mask_img_.get_data()) # grid of locations
x_inds, y_inds, z_inds = inds # unpack tuple
list_niis = []
for inii in xrange(n_random_niis):
# if inii % 24 == 0:
# print "%i/%i" % (inii + 1, n_random_niis)
nfoci = np.random.randint(*n_random_foci)
cur_img = np.zeros(masker.mask_img_.shape)
for ifocus in xrange(nfoci):
# find random point within mask
i = np.random.randint(0, len(x_inds) - 1)
x, y, z = x_inds[i], y_inds[i], z_inds[i]
# put a dot there
if np.random.randint(0, 2):
cur_img[x, y, z] = 150
else:
cur_img[x, y, z] = -150
# smooth current image of random foci
cur_fwhm = np.random.randint(*fwhm_range)
new_nii = smooth_img(
nib.Nifti1Image(cur_img, masker.mask_img_.get_affine(),
header=masker.mask_img_.get_header()), cur_fwhm)
list_niis.append(new_nii)
X_rest = masker.transform(list_niis)
return np.vstack((X_rest, X_rest, X_rest, X_rest,
X_rest, X_rest, X_rest, X_rest))
from nilearn import datasets, plotting, image
data = datasets.fetch_adhd()
mean_func = image.mean_img(data.func[0])
for smoothing in range(0, 25, 5):
plotting.plot_epi(image.smooth_img(mean_func, smoothing),
title="Smoothing %imm" % smoothing)
As we vary the smoothing FWHM, note how we decrease the amount of noise,
but also loose spatial details. In general, the best amount of smoothing
for a given analysis depends on the spatial extent of the effects that
are expected.
"""
from nilearn import datasets, plotting, image
data = datasets.fetch_development_fmri(n_subjects=1)
# Print basic information on the dataset
print('First subject functional nifti image (4D) are located at: %s' %
data.func[0])
first_epi_file = data.func[0]
# First the compute the mean image, from the 4D series of image
mean_func = image.mean_img(first_epi_file)
# Then we smooth, with a varying amount of smoothing, from none to 20mm
# by increments of 5mm
for smoothing in range(0, 25, 5):
smoothed_img = image.smooth_img(mean_func, smoothing)
plotting.plot_epi(smoothed_img,
title="Smoothing %imm" % smoothing)
plotting.show()
subject_IDs = [str(i).zfill(3) for i in range(1, 17)]
IDs = list(zip([run_ID for _ in range(16) for run_ID in run_IDs],
[subject_ID for _ in range(3) for subject_ID in subject_IDs]))
IDs.sort()
# Do this for all subjects/runs:
for ID in IDs:
run, subject = ID
# Extract necessary data
sub = ds005(subject, run)
data = sub.filtered.data
affine = sub.filtered.affine
img = sub.filtered.img
smooth_img = image.smooth_img(img, 6) # smoothed image
# Save paths for input and output
path_result = "results/run%s/visualization/sub%s/" % ID
file_path = "results/run%s/glm/sub%s/" % ID
try:
os.makedirs(path_result)
except OSError:
if not os.path.isdir(path_result):
raise
t_map_gain_file = file_path + "t_stat_gain.nii.gz"
t_map_loss_file = file_path + "t_stat_loss.nii.gz"
t_map_dist_file = file_path + "t_stat_dist2indiff.nii.gz"
n_samples, n_features = images_masked.shape
print n_samples, "subjects, ", n_features, "features"
### Perform massively univariate analysis with permuted OLS ###################
print "Massively univariate model"
neg_log_pvals, all_scores, _ = permuted_ols(
cdr, images_masked, age, # + intercept as a covariate by default
n_perm=1000,
n_jobs=-1) # can be changed to use more CPUs
neg_log_pvals_unmasked = nifti_masker.inverse_transform(
neg_log_pvals).get_data()[..., 0]
### Show results
print "Plotting results"
# background anat
mean_anat = smooth_img(images[0], FWHM).get_data()
for img in images[1:]:
mean_anat += smooth_img(img, FWHM).get_data()
mean_anat /= float(len(images))
ref_img = nibabel.load(images[0])
picked_slice = 36
vmin = -np.log10(0.1) # 10% corrected
plt.figure(figsize=(5, 4))
p_ma = np.ma.masked_less(neg_log_pvals_unmasked, vmin)
plt.imshow(np.rot90(mean_anat[..., picked_slice]),
interpolation='nearest', cmap=plt.cm.gray, vmin=0., vmax=1.)
im = plt.imshow(np.rot90(p_ma[..., picked_slice]),
interpolation='nearest', cmap=plt.cm.autumn,
vmin=vmin, vmax=np.amax(neg_log_pvals_unmasked))
plt.axis('off')
plt.colorbar(im)
# Import the os module, for file manipulation
import os
#########################################################################
# Let us use a Nifti file that is shipped with nilearn
from nilearn.datasets import data
anat_filename = os.path.join(os.path.dirname(data.__file__),
'avg152T1_brain.nii.gz')
print('anat_filename: %s' % anat_filename)
#########################################################################
# Using simple image nilearn functions
from nilearn import image
# functions containing 'img' can take either a filename or an image as input
smooth_anat_img = image.smooth_img(anat_filename, 3)
# While we are giving a file name as input, the object that is returned
# is a 'nibabel' object. It has data, and an affine
anat_data = smooth_anat_img.get_data()
print('anat_data has shape: %s' % str(anat_data.shape))
anat_affine = smooth_anat_img.get_affine()
print('anat_affineaffine:\n%s' % anat_affine)
# Finally, it can be passed to nilearn function
smooth_anat_img = image.smooth_img(smooth_anat_img, 3)
#########################################################################
# Visualization
from nilearn import plotting
cut_coords = (0, 0, 0)
from nilearn.input_data import MultiNiftiMasker
from scipy.stats import norm
from statsmodels.sandbox.stats.multicomp import fdrcorrection0
from nilearn.plotting import plot_glass_brain
from scipy.stats import ttest_1samp
import sys
from nibabel import load
from nilearn.image import smooth_img
models = sys.argv[1:]
mask = 'brainmask_group_template.nii.gz'
valid_subjs = [1,2,5,6,7,8,9,11,12,14,15,16,17,18,19]
scores_bsc = np.arctanh(apply_mask(smooth_img(sorted(glob.glob('MaThe/avg_maps/model_{}_*whole*'.format(models[0]))), 3.0), mask_img=mask)).astype('float32')
scores_mfs = np.arctanh(apply_mask(smooth_img(sorted(glob.glob('MaThe/avg_maps/model_{}_*whole*'.format(models[1]))), 3.0), mask_img=mask)).astype('float32')
scores_bsc[scores_bsc<0] = 0
scores_mfs[scores_mfs<0] = 0
diff_scores = scores_bsc - scores_mfs
mean_diff = diff_scores.mean(axis=0)
tscores, p_values = ttest_1samp(diff_scores, 0, axis=0)
p_values[np.isnan(p_values)] = 1
which_ones = p_values > 0.05
mean_diff[which_ones] = 0
tscores[which_ones] = 0
display = fsf.plot_diff_avg_whole(mean_diff, 0.001)
display.savefig('mean_unthresh_diff_model_whole_3mm_{}.svg'.format('_'.join(models)))
display.savefig('mean_unthresh_diff_model_whole_3mm_{}.png'.format('_'.join(models)))
fsf.save_map_avg_whole(mean_diff, threshold=None, model='3mm_diff_'+'_'.join(models))
display = fsf.plot_diff_avg_whole(tscores, 0.001)
# Block Design using original data
start = np.loadtxt('../../../data/convo/' + f1 + '_conv001.txt')[4:]
end = np.loadtxt('../../../data/convo/' + f1 + '_conv004.txt')[4:]
convo = np.loadtxt('../../../data/convo/' + f1 + '_conv005.txt')[4:]
# Building design X matrix
design = np.ones((len(convo), 4))
design[:, 1] = start
design[:, 2] = end
design[:, 3] = convo
X = design
beta, errors, RSS, df = linear_modeling.beta_est(dat, X)
pval = normal_assumption.sw(errors)
# smoothing
fmri_img = image.smooth_img('../../../data/sub001/BOLD/' + f1 + '/filtered_func_data_mni.nii.gz', fwhm=6)
mean_img = image.mean_img(fmri_img)
# Thresholding
p_val = np.ones(vol_shape + (pval.shape[1],))
p_val[in_brain_mask, :] = pval
log_p_values = -np.log10(p_val[..., 0])
log_p_values[np.isnan(log_p_values)] = 0.
log_p_values[log_p_values > 10.] = 10.
log_p_values[log_p_values < -np.log10(0.05/137)] = 0
plot_stat_map(nib.Nifti1Image(log_p_values, img.get_affine()),
mean_img, title='SW Test p-values for Block Design, original',
annotate=False,colorbar=True)
plt.savefig("../../../data/GLS/block_normality_test.png")
#--------------------------------------------------------------------------
# Block Design using smoothed data
mask_img = load_mni152_brain_mask()
masker = NiftiMasker(
mask_img=mask_img, memory='nilearn_cache', memory_level=1)
masker = masker.fit()
# Images may fail to be transformed, and are of different shapes,
# so we need to transform one-by-one and keep track of failures.
X = []
is_usable = np.ones((len(images),), dtype=bool)
for index, image_path in enumerate(images):
# load image and remove nan and inf values.
# applying smooth_img to an image with fwhm=None simply cleans up
# non-finite values but otherwise doesn't modify the image.
image = smooth_img(image_path, fwhm=None)
try:
X.append(masker.transform(image))
except Exception as e:
meta = nv_data['images_meta'][index]
print("Failed to mask/reshape image: id: {0}; "
"name: '{1}'; collection: {2}; error: {3}".format(
meta.get('id'), meta.get('name'),
meta.get('collection_id'), e))
is_usable[index] = False
# Now reshape list into 2D matrix, and remove failed images from terms
X = np.vstack(X)
term_weights = term_weights[is_usable, :]
# MNI152_FILE_PATH is nothing more than a string with a path pointing to # a nifti image. You can replace it with a string pointing to a file on # your disk. Note that it should be a 3D volume, and not a 4D volume. ######################################################################### # Simple image manipulation: smoothing # ------------------------------------- # # Let's use an image-smoothing function from nilearn: # :func:`nilearn.image.smooth_img` # # Functions containing 'img' can take either a filename or an image as input. # # Here we give as inputs the image filename and the smoothing value in mm from nilearn import image smooth_anat_img = image.smooth_img(MNI152_FILE_PATH, fwhm=3) # While we are giving a file name as input, the function returns # an in-memory object: print(smooth_anat_img) ######################################################################### # This is an in-memory object. We can pass it to nilearn function, for # instance to look at it plotting.plot_img(smooth_anat_img) ######################################################################### # We could also pass it to the smoothing function more_smooth_anat_img = image.smooth_img(smooth_anat_img, fwhm=3) plotting.plot_img(more_smooth_anat_img)
from nilearn.masking import apply_mask
from nilearn.input_data import MultiNiftiMasker
from scipy.stats import norm
from statsmodels.sandbox.stats.multicomp import fdrcorrection0
from nilearn.plotting import plot_glass_brain
from nilearn.image import smooth_img
from scipy.stats import ttest_1samp
import sys
from nibabel import load
model = sys.argv[1]
mask = 'brainmask_group_template.nii.gz'
valid_subjs = [1,2,5,6,7,8,9,11,12,14,15,16,17,18,19]
scores = apply_mask(smooth_img(sorted(glob.glob('MaThe/avg_maps/model_depcor_{}_*'.format(model))), 5.0), mask_img=mask).astype('float32')
mean_scores = scores.mean(axis=0)
tscores, p_values = ttest_1samp(scores, 0, axis=0)
p_values[np.isnan(p_values)] = 1
which_ones = p_values > 0.05
mean_scores[which_ones] = 0
tscores[which_ones] = 0
display = fsf.plot_diff_avg_whole(mean_scores, 0.001)
display.savefig('mean_5mm_unthresh_depcor_model_whole_{}.svg'.format(model))
display.savefig('mean_5mm_unthresh_depcor_model_whole_{}.png'.format(model))
fsf.save_map_avg_whole(mean_scores, threshold=None, model='5mm_depcor_'+model)
display = fsf.plot_diff_avg_whole(tscores, 0.001)
display.savefig('ttest_5mm_unthresh_depcor_model_whole_{}.svg'.format(model))
display.savefig('ttest_5mm_unthresh_depcor_model_whole_{}.png'.format(model))
import numpy as np
import featurespace_fun as fsf
import matplotlib.pyplot as plt
from nilearn.masking import apply_mask
from nilearn.image import smooth_img
from scipy.stats import norm
from statsmodels.sandbox.stats.multicomp import fdrcorrection0
from scipy.stats import ttest_1samp
import sys
from scipy.stats.mstats import trimmed_mean_ci
from scipy.stats import ttest_1samp, trim_mean
models = sys.argv[1:]
#models = ['logBSC_H200_ds_conv', 'logMFS_ds']
mask = 'brainmask_group_template.nii.gz'
scores_bsc = np.arctanh(apply_mask(smooth_img(glob.glob('MaThe/avg_maps/model_{}_*whole*'.format(models[0])), fwhm=3.0), mask_img=mask))
scores_mfs = np.arctanh(apply_mask(smooth_img(glob.glob('MaThe/avg_maps/model_{}_*whole*'.format(models[1])), fwhm=3.0), mask_img=mask))
diff_scores = scores_bsc - scores_mfs
mean_diff = trim_mean(diff_scores, 0.08, axis=0)
trim_mean_ci = trimmed_mean_ci(diff_scores, (0.08, 0.08), axis=0)
which_ones = np.logical_not(np.logical_or(trim_mean_ci[0,:] > 0, trim_mean_ci[1,:] < 0))
mean_diff[which_ones] = 0
display = fsf.plot_diff_avg_whole(mean_diff, 0.001)
display.savefig('mean_diff_smoothed_trim_model_{}.svg'.format('_'.join(models)))
display.savefig('mean_diff_smoothed_trim_model_{}.png'.format('_'.join(models)))
fsf.save_map_avg_whole(mean_diff, threshold=None, model='diff_smooth_trim_'+'_'.join(models))
cax = fig.add_axes([0.9, 0.15, 0.03, 0.7])
fig.colorbar(im, cax=cax)
plt.show()
# Visualizing p values for the middle slice in gray
fig, axes = plt.subplots(nrows=3, ncols=4)
for i, ax in zip(range(0,design_mat_filter.shape[1] - 1,1), axes.flat):
im = ax.imshow(p_val[:, :, 45, i], cmap = 'gray')
fig.subplots_adjust(right=0.85)
cax = fig.add_axes([0.9, 0.15, 0.03, 0.7])
fig.colorbar(im, cax=cax)
plt.show()
fmri_img = image.smooth_img('../../../data/sub001/BOLD/task001_run001/filtered_func_data_mni.nii.gz', fwhm=6)
mean_img = image.mean_img(fmri_img)
# Thresholding
p_val = np.ones(vol_shape + (p_mat.shape[0],))
p_val[in_brain_mask, :] = p_mat.T
log_p_values = -np.log10(p_val[..., 6])
log_p_values[np.isnan(log_p_values)] = 0.
log_p_values[log_p_values > 10.] = 10.
log_p_values[log_p_values < -np.log10(0.05/137)] = 0
plot_stat_map(nib.Nifti1Image(log_p_values, img.get_affine()),
mean_img, title='Thresholded p-values after Filtering (beta 6)', annotate=False,
colorbar=True)
################### Noise from our model
# voxel time course
print('anat_filename: %s' % anat_filename)
# Using nibabel.load to load existing Nifti image #############################
import nibabel
anat_img = nibabel.load(anat_filename)
# Accessing image data and affine #############################################
anat_data = anat_img.get_data()
print('anat_data has shape: %s' % str(anat_data.shape))
anat_affine = anat_img.get_affine()
print('anat_affine:\n%s' % anat_affine)
# Using image in nilearn functions ############################################
from nilearn import image
# functions containing 'img' can take either a filename or an image as input
smooth_anat_img = image.smooth_img(anat_filename, 6)
smooth_anat_img = image.smooth_img(anat_img, 6)
# Visualization ###############################################################
from nilearn import plotting
cut_coords = (0, 0, 0)
plotting.plot_anat(anat_filename, cut_coords=cut_coords,
title='Anatomy image')
plotting.plot_anat(smooth_anat_img,
cut_coords=cut_coords,
title='Smoothed anatomy image')
# Saving image to file ########################################################
smooth_anat_img.to_filename('smooth_anat_img.nii.gz')
# When using methods that are not robust to noise, it is useful to apply a # spatial filtering kernel on the data. Such data smoothing is usually applied # using a Gaussian function with 4mm to 12mm full-width at half-maximum (this # is where the FWHM comes from). The function :func:`nilearn.image.smooth_img` # accounts for potential anisotropy in the image affine (i.e., non-indentical # voxel size in all the three dimensions). Analogous to the majority of nilearn # functions, smooth_img function can also use file names as input parameters. # Smooth the data using image processing module from nilearn from nilearn import image # Functional data fmri_filename = haxby_dataset.func[0] # smoothing: first argument as functional data filename and smoothing value # (integer) in second argument. Output returns in Nifti image. fmri_img = image.smooth_img(fmri_filename, fwhm=6) # Visualize the mean of the smoothed EPI image using plotting function # `plot_epi` from nilearn.plotting import plot_epi # First, compute the voxel-wise mean of smooth EPI image (first argument) using # image processing module `image` mean_img = image.mean_img(fmri_img) # Second, we visualize the mean image with coordinates positioned manually plot_epi(mean_img, title='Smoothed mean EPI', cut_coords=cut_coords) ############################################################################## # Given the smoothed functional data stored in variable 'fmri_img', we then # select two features of interest with face and house experimental conditions. # The method we will be using is a simple Student's t-test. The below section
def do_subject_glm(subject_id):
subject_output_dir = os.path.join(output_dir, subject_id)
# make design matrices
design_matrices = []
func = []
anat = os.path.join(subject_output_dir, "anatomy", "whighres001_brain.nii")
for run_path in sorted(glob.glob(os.path.join(
data_dir, subject_id, "model/model001/onsets/task*"))):
run_id = os.path.basename(run_path)
run_func = glob.glob(os.path.join(subject_output_dir, "BOLD", run_id,
"wrbold*.nii"))
assert len(run_func) == 1
run_func = run_func[0]
run_onset_paths = sorted(glob.glob(os.path.join(
data_dir, subject_id, "model/model001/onsets/%s/*" % run_id)))
onsets = map(np.loadtxt, run_onset_paths)
conditions = np.hstack(
[[condition_keys["cond%03i" % (c + 1)]] * len(onsets[c])
for c in range(len(run_onset_paths))])
onsets = np.vstack((onsets))
onsets *= tr
run_func = nibabel.load(run_func)
func.append(run_func)
n_scans = run_func.shape[-1]
onset, duration, modulation = onsets.T
frametimes = np.linspace(0, (n_scans - 1) * tr, n_scans)
paradigm = pd.DataFrame(dict(name=conditions, onset=onset,
duration=duration, modulation=modulation))
design_matrix = make_design_matrix(frametimes, paradigm,
hrf_model=hrf_model,
drift_model=drift_model,
period_cut=hfcut)
design_matrices.append(design_matrix)
n_runs = len(func)
# specify contrasts
_, _, names = check_design_matrix(design_matrix)
n_columns = len(names)
contrast_matrix = np.eye(n_columns)
contrasts = {}
for c in range(len(condition_keys)):
contrasts[names[2 * c]] = contrast_matrix[2 * c]
contrasts["avg"] = np.mean(contrasts.values(), axis=0)
# more interesting contrasts
contrasts_ = {}
for contrast, val in contrasts.items():
if not contrast == "avg":
contrasts_["%s_minus_avg" % contrast] = val - contrasts["avg"]
contrasts = contrasts_
# fit GLM
from nilearn.image import smooth_img
func = smooth_img(func, fwhm=8.)
print 'Fitting a GLM (this takes time)...'
fmri_glm = FMRILinearModel(func, [check_design_matrix(design_matrix)[1]
for design_matrix in design_matrices],
mask='compute')
fmri_glm.fit(do_scaling=True, model='ar1')
# save computed mask
mask_path = os.path.join(subject_output_dir, "mask.nii")
print "Saving mask image to %s ..." % mask_path
nibabel.save(fmri_glm.mask, mask_path)
# compute contrast maps
z_maps = {}
effects_maps = {}
for contrast_id, contrast_val in contrasts.items():
print "\tcontrast id: %s" % contrast_id
z_map, t_map, effects_map, var_map = fmri_glm.contrast(
[contrast_val] * n_runs, con_id=contrast_id, output_z=True,
output_stat=True, output_effects=True, output_variance=True)
for map_type, out_map in zip(['z', 't', 'effects', 'variance'],
[z_map, t_map, effects_map, var_map]):
map_dir = os.path.join(subject_output_dir, '%s_maps' % map_type)
if not os.path.exists(map_dir):
os.makedirs(map_dir)
map_path = os.path.join(
map_dir, '%s.nii.gz' % contrast_id)
print "\t\tWriting %s ..." % map_path
nibabel.save(out_map, map_path)
if map_type == 'z':
z_maps[contrast_id] = map_path
if map_type == 'effects':
effects_maps[contrast_id] = map_path
# # generate stats report
# stats_report_filename = os.path.join(subject_output_dir, "reports",
# "report_stats.html")
# generate_subject_stats_report(
# stats_report_filename, contrasts, z_maps, fmri_glm.mask, anat=anat,
# threshold=2.3, cluster_th=15, design_matrices=design_matrices, TR=tr,
# subject_id="sub001", n_scans=n_scans, hfcut=hfcut,
# paradigm=paradigm, frametimes=frametimes,
# drift_model=drift_model, hrf_model=hrf_model)
# ProgressReport().finish_dir(subject_output_dir)
return dict(subject_id=subject_id, mask=mask_path,
effects_maps=effects_maps, z_maps=z_maps, contrasts=contrasts)
brain_mask = sys.argv[2]
gm_mask = sys.argv[3]
max_lag = int(sys.argv[4])
out_dir = sys.argv[5]
out_prefix = sys.argv[6]
smooth_kernel = sys.argv[7]
lags = range(-max_lag, max_lag+1)
brain_mask_img = nib.load(brain_mask)
brain_mask_data = brain_mask_img.get_data().astype(bool)
gm_mask_img = nib.load(gm_mask)
gm_mask_data = gm_mask_img.get_data().astype(bool)
epi_img = nib.load(epi)
epi_img_resampled = image.resample_img(epi_img, target_affine=brain_mask_img.get_affine(), target_shape=brain_mask_img.shape)
epi_data_resampled = epi_img_resampled.get_data()
lag_map_data = gen_lag_map(epi_data_resampled, brain_mask_data, gm_mask_data, lags)
lag_map_image = masking.unmask(lag_map_data, brain_mask_img)
lag_map_image_smoothed = image.smooth_img(lag_map_image, float(smooth_kernel))
nib.save(lag_map_image, os.path.join(out_dir, out_prefix+'_lag{}.nii.gz'.format(str(max_lag))))
nib.save(lag_map_image_smoothed, os.path.join(out_dir, out_prefix+'_lag{}_smoothed{}.nii.gz'.format(str(max_lag),smooth_kernel)))
Conv3D(16, kernel_size=3, activation='relu', padding="same"),
BatchNormalization(),
Conv3DTranspose(16, kernel_size=3, strides=2, activation='relu',
padding="same"),
BatchNormalization(),
Conv3D(1, kernel_size=3, activation=None, padding="same"),
], name="decoder")
autoencoder = Sequential([encoder, decoder], name="autoencoder")
return encoder, decoder, autoencoder
if __name__ == "__main__":
data = datasets.fetch_haxby(subjects=(2,))
fmri_filename = data.func[0]
smoothed_img = image.smooth_img(fmri_filename, 2)
smoothed_data = smoothed_img.get_data().transpose(3, 0, 1, 2)
#mean = smoothed_data.mean(axis=0)
#smoothed_data -= mean
#scale = smoothed_data.std(axis=0) + 1e-6
scale = smoothed_data.std() # global scale
smoothed_data /= scale
smoothed_data = smoothed_data[:, :, :, :, None]
input_shape = smoothed_data.shape[1:]
smoothed_data_train = smoothed_data[:1200]
smoothed_data_test = smoothed_data[1200:]
encoder, decoder, autoencoder = make_models(input_shape=input_shape)
autoencoder.compile(optimizer=Adam(lr=0.001), loss="mse")
haxby_labels = np.genfromtxt(haxby_files.session_target[0], skip_header=1,
usecols=[0], dtype=basestring)
### Visualization function ####################################################
import matplotlib.pyplot as plt
from nilearn.plotting import plot_epi, plot_stat_map, plot_roi
from nilearn.input_data import NiftiLabelsMasker
plt.close('all')
### Find voxels of interest ###################################################
# Smooth the data
from nilearn import image
fmri_img = image.smooth_img(haxby_files.func[0], fwhm=6)
# Plot the mean image
fig_id = plt.subplot(2, 1, 1)
mean_img = image.mean_img(fmri_img)
plot_epi(mean_img, title='Smoothed mean EPI', cut_coords=(coronal, sagittal,
axial), axes=fig_id)
# Run a T-test for face and houses
from scipy import stats
fmri_data = fmri_img.get_data()
_, p_values = stats.ttest_ind(fmri_data[..., haxby_labels == 'face'],
fmri_data[..., haxby_labels == 'house'],
axis=-1)
# Use a log scale for p-values