def _check_vol_to_surf_results(img, mesh):
mni_mask = datasets.load_mni152_brain_mask()
for kind, interpolation, mask_img in itertools.product(
['ball', 'line'], ['linear', 'nearest'], [mni_mask, None]):
proj_1 = vol_to_surf(img,
mesh,
kind=kind,
interpolation=interpolation,
mask_img=mask_img)
assert_true(proj_1.ndim == 1)
img_rot = image.resample_img(img,
target_affine=rotation(
np.pi / 3., np.pi / 4.))
proj_2 = vol_to_surf(img_rot,
mesh,
kind=kind,
interpolation=interpolation,
mask_img=mask_img)
# The projection values for the rotated image should be close
# to the projection for the original image
diff = np.abs(proj_1 - proj_2) / np.abs(proj_1)
assert_true(np.mean(diff[diff < np.inf]) < .03)
img_4d = image.concat_imgs([img, img])
proj_4d = vol_to_surf(img_4d,
mesh,
kind=kind,
interpolation=interpolation,
mask_img=mask_img)
nodes, _ = surface.load_surf_mesh(mesh)
assert_array_equal(proj_4d.shape, [nodes.shape[0], 2])
assert_array_almost_equal(proj_4d[:, 0], proj_1, 3)
def save_imgs(bg_img, stats_img, seed_img, output_path):
vmin, vmax = abs(args.vmin), abs(args.vmax)
if args.mask:
stats_img = image.math_img('img1 * img2',
img1=stats_img,
img2=datasets.load_mni152_brain_mask())
slices = {'x': (-71, 72), 'y': (-107, 74), 'z': (-70, 82)}
for key, value in slices.items():
for i in range(value[0], value[1]):
if args.verbose: print(f'slice: {key}={i}')
img = plotting.plot_stat_map(bg_img=bg_img,
stat_map_img=stats_img,
threshold=vmin,
vmax=vmax,
display_mode=key,
cut_coords=[i],
colorbar=False,
annotate=False,
draw_cross=False)
if args.seed: img.add_overlay(seed_img, cmap='cold_hot', alpha=.25)
img.savefig(f'{output_path}_{key}={i}.png', dpi=600)
img.close()
del img
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 get_template(space='mni152_1mm', mask=None):
if space == 'mni152_1mm':
if mask is None:
img = nib.load(datasets.fetch_icbm152_2009()['t1'])
elif mask == 'brain':
img = nib.load(datasets.fetch_icbm152_2009()['mask'])
elif mask == 'gm':
img = datasets.fetch_icbm152_brain_gm_mask(threshold=0.2)
else:
raise ValueError('Mask {0} not supported'.format(mask))
elif space == 'mni152_2mm':
if mask is None:
img = datasets.load_mni152_template()
elif mask == 'brain':
img = datasets.load_mni152_brain_mask()
elif mask == 'gm':
# this approach seems to approximate the 0.2 thresholded
# GM mask pretty well
temp_img = datasets.load_mni152_template()
data = temp_img.get_data()
data = data * -1
data[data != 0] += np.abs(np.min(data))
data = (data > 1200).astype(int)
img = nib.Nifti1Image(data, temp_img.affine)
else:
raise ValueError('Mask {0} not supported'.format(mask))
else:
raise ValueError('Space {0} not supported'.format(space))
return img
def find_overlaps(atlas_names, dimension):
"""Estimate overlap (queries) provided atlas names
Parameters
----------
atlas_names : str or list of str
Grab atlas from web given the name. Few are shipped with FSL
and Nilearn.
Valid options: ['harvard_oxford', 'destrieux', 'diedrichsen',
'juelich', 'jhu', 'mist', 'yeo_networks7',
'yeo_networks17']
dimension : int
Dimension of the DiFuMo atlas (4D).
Returns
-------
info : dict
Contains meta-data assigned to each atlas name such as
overlap proportion, etc
Each atlas dict contains following attributes:
'intersection' : sparse matrix
dot product between DiFuMo regions and regions in target
atlas (existing pre-defined)
'target_size' : np.ndarray
Size of each region estimated in target atlas
'overlap_proportion' : list of pd.Series
Each list contains the proportion of overlap estimated
between this region and all region in target sizes.
Sorted according to most strong hit in the overlap.
'overlap_size' : list
Each list contain overlap in estimated sizes for all
regions in target atlas.
save_labels : dict
List of original labels specific to each atlas. Useful
for matching the overlap for the visualization of
records.
"""
masker = input_data.NiftiMasker(datasets.load_mni152_brain_mask()).fit([])
queries = transform_difumo_to_data(dimension=dimension, masker=masker)
atlases = fetch_atlases(atlas_names)
info = {}
save_labels = {}
for atlas in atlas_names:
this_atlas = atlases[atlas]
labels_img, labels = labels_img_to_binary(this_atlas.maps,
this_atlas.labels, masker)
save_labels[atlas] = labels
info[atlas] = overlaps(queries, labels_img)
return info, save_labels
def coords_to_voxels(coords, ref_img=None):
if ref_img is None:
ref_img = load_mni152_brain_mask()
affine = ref_img.affine
coords = np.atleast_2d(coords)
coords = np.hstack([coords, np.ones((len(coords), 1))])
voxels = np.linalg.pinv(affine).dot(coords.T)[:-1].T
voxels = voxels[(voxels >= 0).all(axis=1)]
voxels = voxels[(voxels < ref_img.shape[:3]).all(axis=1)]
voxels = np.floor(voxels).astype(int)
return voxels
def get_template(space='mni152_1mm', mask=None):
"""
Load template file.
Parameters
----------
space : {'mni152_1mm', 'mni152_2mm', 'ale_2mm'}, optional
Template to load. Default is 'mni152_1mm'.
mask : {None, 'brain', 'gm'}, optional
Whether to return the raw template (None), a brain mask ('brain'), or
a gray-matter mask ('gm'). Default is None.
Returns
-------
img : :obj:`nibabel.nifti1.Nifti1Image`
Template image object.
"""
if space == 'mni152_1mm':
if mask is None:
img = nib.load(datasets.fetch_icbm152_2009()['t1'])
elif mask == 'brain':
img = nib.load(datasets.fetch_icbm152_2009()['mask'])
elif mask == 'gm':
img = datasets.fetch_icbm152_brain_gm_mask(threshold=0.2)
else:
raise ValueError('Mask {0} not supported'.format(mask))
elif space == 'mni152_2mm':
if mask is None:
img = datasets.load_mni152_template()
elif mask == 'brain':
img = datasets.load_mni152_brain_mask()
elif mask == 'gm':
# this approach seems to approximate the 0.2 thresholded
# GM mask pretty well
temp_img = datasets.load_mni152_template()
data = temp_img.get_data()
data = data * -1
data[data != 0] += np.abs(np.min(data))
data = (data > 1200).astype(int)
img = nib.Nifti1Image(data, temp_img.affine)
else:
raise ValueError('Mask {0} not supported'.format(mask))
elif space == 'ale_2mm':
if mask is None:
img = datasets.load_mni152_template()
else:
# Not the same as the nilearn brain mask, but should correspond to
# the default "more conservative" MNI152 mask in GingerALE.
img = nib.load(
op.join(get_resource_path(),
'templates/MNI152_2x2x2_brainmask.nii.gz'))
else:
raise ValueError('Space {0} not supported'.format(space))
return img
def test_kernel_peaks(testdata_cbma, tmp_path_factory, kern, res, param,
return_type, kwargs):
"""Peak/COMs of kernel maps should match the foci fed in (assuming focus isn't masked out).
Notes
-----
Remember that dataframe --> dataset won't work.
Only testing dataset --> dataset with ALEKernel because it takes a while.
Test on multiple template resolutions.
"""
tmpdir = tmp_path_factory.mktemp("test_kernel_peaks")
testdata_cbma.update_path(tmpdir)
id_ = "pain_03.nidm-1"
# Ignoring resolution until we support 0.8.1
# template = load_mni152_brain_mask(resolution=res)
template = load_mni152_brain_mask()
masker = get_masker(template)
xyz = testdata_cbma.coordinates.loc[testdata_cbma.coordinates["id"] == id_,
["x", "y", "z"]]
ijk = mm2vox(xyz, masker.mask_img.affine)
ijk = np.squeeze(ijk.astype(int))
if param == "dataframe":
input_ = testdata_cbma.coordinates.copy()
elif param == "dataset":
input_ = testdata_cbma.copy()
kern_instance = kern(**kwargs)
output = kern_instance.transform(input_, masker, return_type=return_type)
if return_type == "image":
kern_data = output[0].get_fdata()
elif return_type == "array":
kern_data = np.squeeze(
masker.inverse_transform(output[:1, :]).get_fdata())
else:
f = output.images.loc[output.images["id"] == id_,
kern_instance.image_type].values[0]
kern_data = nib.load(f).get_fdata()
if isinstance(kern_instance, kernel.ALEKernel):
loc_idx = np.array(np.where(kern_data == np.max(kern_data))).T
elif isinstance(kern_instance, (kernel.MKDAKernel, kernel.KDAKernel)):
loc_idx = np.array(center_of_mass(kern_data)).astype(int).T
else:
raise Exception(f"A {type(kern_instance)}? Why?")
loc_ijk = np.squeeze(loc_idx)
assert np.array_equal(ijk, loc_ijk)
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 fishers_workflow(img_list=None, prefix=None, output_dir=None):
from nimare.meta.esma import fishers
from nilearn.datasets import load_mni152_brain_mask
import nibabel as nib
import numpy as np
import os.path as op
import statsmodels.stats.multitest as mc
mask_img = load_mni152_brain_mask()
mask_ind = np.nonzero(mask_img.get_data())
img_stack = []
for i, img_fn in enumerate(img_list):
tmp_img = nib.load(img_fn)
if i == 0:
img_stack = tmp_img.get_data()[mask_ind]
else:
img_stack = np.vstack([img_stack, tmp_img.get_data()[mask_ind]])
results = fishers(img_stack, two_sided=False)
for tmp_key in results.keys():
img_data = np.zeros(mask_img.shape)
img_data[mask_ind] = results[tmp_key]
img = nib.Nifti1Image(img_data, mask_img.affine)
nib.save(
img,
op.join(
output_dir, '{prefix}_{suffix}.nii.gz'.format(prefix=prefix,
suffix=tmp_key)))
_, p_corr = mc.fdrcorrection(results['p'],
alpha=0.05,
method='indep',
is_sorted=False)
img_data = np.zeros(mask_img.shape)
img_data[mask_ind] = p_corr
img = nib.Nifti1Image(img_data, mask_img.affine)
nib.save(
img,
op.join(output_dir,
'{prefix}_p_corr-fdr05.nii.gz'.format(prefix=prefix)))
def generate_mni_space_img(n_scans=1, res=30, random_state=0, mask_dilation=2):
rng = check_random_state(random_state)
mni = datasets.load_mni152_brain_mask()
target_affine = np.eye(3) * res
mask_img = image.resample_img(
mni, target_affine=target_affine, interpolation="nearest")
masker = input_data.NiftiMasker(mask_img).fit()
n_voxels = image.get_data(mask_img).sum()
data = rng.randn(n_scans, n_voxels)
if mask_dilation is not None and mask_dilation > 0:
mask_img = image.new_img_like(
mask_img, ndimage.binary_dilation(
image.get_data(mask_img), iterations=mask_dilation))
return masker.inverse_transform(data), mask_img
def __create_precomputed(data_dir: Optional[Union[str, Path]] = None) -> None:
"""Create nearest neighbor interpolation niftis for MATLAB."""
# Embed import to prevent circular dependency.
from brainstat.datasets import fetch_template_surface
data_dir = Path(
data_dir) if data_dir else data_directories["BRAINSTAT_DATA_DIR"]
mni152 = load_mni152_brain_mask()
for template in ("fsaverage5", "fsaverage", "civet41k", "civet164k"):
output_file = data_dir / f"nn_interp_{template}.nii.gz"
if output_file.exists():
continue
top_surf = "pial" if template[:9] == "fsaverage" else "mid"
pial = fetch_template_surface(template, layer=top_surf, join=False)
white = fetch_template_surface(template, layer="white", join=False)
labels = (
np.arange(1,
get_points(pial[0]).shape[0] + 1),
np.arange(
get_points(pial[0]).shape[0] + 1,
get_points(pial[0]).shape[0] * 2 + 1),
)
multi_surface_to_volume(
pial=pial,
white=white,
volume_template=mni152,
labels=labels,
output_file=str(output_file),
interpolation="nearest",
)
if not (data_dir / "nn_interp_hcp.nii.gz").exists():
import hcp_utils as hcp
from brainspace.mesh.mesh_creation import build_polydata
pial_fslr32k = (build_polydata(hcp.mesh.pial[0], hcp.mesh.pial[1]), )
white_fslr32k = (build_polydata(hcp.mesh.white[0],
hcp.mesh.white[1]), )
labels_fslr32k = (np.arange(1,
get_points(pial_fslr32k[0]).shape[0] +
1), )
multi_surface_to_volume(
pial=pial_fslr32k,
white=white_fslr32k,
volume_template=mni152,
labels=labels_fslr32k,
output_file=str(data_dir / "nn_interp_fslr32k.nii.gz"),
interpolation="nearest",
)
def get_masker(mask_img=None, target_affine=None):
if isinstance(mask_img, input_data.NiftiMasker):
return mask_img
if mask_img is None:
mask_img = load_mni152_brain_mask()
if target_affine is not None:
if np.ndim(target_affine) == 0:
target_affine = np.eye(3) * target_affine
elif np.ndim(target_affine) == 1:
target_affine = np.diag(target_affine)
mask_img = image.resample_img(mask_img,
target_affine=target_affine,
interpolation="nearest")
masker = input_data.NiftiMasker(mask_img=mask_img).fit()
return masker
def make_sphere(x, y, z, output_dir):
mask = load_mni152_brain_mask()
mask_img = mask.get_data()
xyz_img = np.dot(np.linalg.inv(mask.affine), [x, y, z, 1]).astype(int)
tmp_mask_img = mask_img * 0
tmp_mask_img[xyz_img[0], xyz_img[1], xyz_img[2]] = 1
tmp_roi_img = nib.Nifti1Image(tmp_mask_img, mask.affine, mask.header)
tmp_roi_fn = op.join(output_dir, '{x}_{y}_{z}.nii.gz'.format(x=x, y=y,
z=z))
nib.save(tmp_roi_img, tmp_roi_fn)
di = fsl.maths.DilateImage(in_file=tmp_roi_fn,
operation="mean",
kernel_shape="sphere",
kernel_size=4,
out_file=tmp_roi_fn)
def find_overlaps(dimension):
"""Estimate overlap (queries) within and across components
Parameters
----------
dimension : int
Dimension of the DiFuMo atlas (4D).
Returns
-------
info : dict
Contains meta-data assigned to each atlas name such as
overlap proportion, etc
Each atlas dict contains following attributes:
'intersection' : sparse matrix
dot product between DiFuMo regions and regions in target
atlas (existing pre-defined)
'target_size' : np.ndarray
Size of each region estimated in target atlas
'overlap_proportion' : list of pd.Series
Each list contains the proportion of overlap estimated
between this region and all region in target sizes.
Sorted according to most strong hit in the overlap.
'overlap_size' : list
Each list contain overlap in estimated sizes for all
regions in target atlas.
"""
masker = input_data.NiftiMasker(datasets.load_mni152_brain_mask()).fit([])
queries = transform_difumo_to_data(dimension=dimension, masker=masker)
info = {}
for n in [64, 128, 256, 512, 1024]:
targets = transform_difumo_to_data(dimension=n, masker=masker)
info[n] = overlaps(queries, targets)
return info
def _check_vol_to_surf_results(img, mesh):
mni_mask = datasets.load_mni152_brain_mask()
for kind, interpolation, mask_img in itertools.product(
['ball', 'line'], ['linear', 'nearest'], [mni_mask, None]):
proj_1 = vol_to_surf(
img, mesh, kind=kind, interpolation=interpolation,
mask_img=mask_img)
assert_true(proj_1.ndim == 1)
img_rot = image.resample_img(
img, target_affine=rotation(np.pi / 3., np.pi / 4.))
proj_2 = vol_to_surf(
img_rot, mesh, kind=kind, interpolation=interpolation,
mask_img=mask_img)
# The projection values for the rotated image should be close
# to the projection for the original image
diff = np.abs(proj_1 - proj_2) / np.abs(proj_1)
assert_true(np.mean(diff[diff < np.inf]) < .03)
img_4d = image.concat_imgs([img, img])
proj_4d = vol_to_surf(
img_4d, mesh, kind=kind, interpolation=interpolation,
mask_img=mask_img)
nodes, _ = surface.load_surf_mesh(mesh)
assert_array_equal(proj_4d.shape, [nodes.shape[0], 2])
assert_array_almost_equal(proj_4d[:, 0], proj_1, 3)
def get_masker(mask_img=None):
if mask_img is None:
mask_img = load_mni152_brain_mask()
masker = input_data.NiftiMasker(mask_img=mask_img).fit()
return masker
def surface_decoder(
pial: Union[str, BSPolyData, Sequence[Union[str, BSPolyData]]],
white: Union[str, BSPolyData, Sequence[Union[str, BSPolyData]]],
stat_labels: Union[str, np.ndarray, Sequence[Union[str, np.ndarray]]],
*,
interpolation: str = "linear",
data_dir: Optional[Union[str, Path]] = None,
database: str = "neurosynth",
) -> pd.DataFrame:
"""Meta-analytic decoding of surface maps using NeuroSynth or NeuroQuery.
Parameters
----------
pial : str, BSPolyData, sequence of str or BSPolyData
Path of a pial surface file, BSPolyData of a pial surface or a list
containing multiple of the aforementioned.
white : str, BSPolyData, sequence of str or BSPolyData
Path of a white matter surface file, BSPolyData of a pial surface or a
list containing multiple of the aforementioned.
stat_labels : str, numpy.ndarray, sequence of str or numpy.ndarray
Path to a label file for the surfaces, numpy array containing the
labels, or a list containing multiple of the aforementioned.
interpolation : str, optional
Either 'nearest' for nearest neighbor interpolation, or 'linear'
for trilinear interpolation, by default 'linear'.
data_dir : str, optional
The directory of the dataset. If none exists, a new dataset will
be downloaded and saved to this path. If None, the directory defaults to
your home directory, by default None.
Returns
-------
pandas.DataFrame
Table with correlation values for each feature.
"""
from nilearn.datasets import load_mni152_brain_mask
data_dir = Path(
data_dir) if data_dir else data_directories["NEUROSYNTH_DATA_DIR"]
data_dir.mkdir(exist_ok=True, parents=True)
logger.info(
"Fetching Neurosynth feature files. This may take several minutes if you haven't downloaded them yet."
)
feature_files = tuple(_fetch_precomputed(data_dir, database=database))
mni152 = load_mni152_brain_mask()
with NamedTemporaryFile(suffix=".nii.gz", delete=False) as f:
name = f.name
try:
multi_surface_to_volume(
pial=pial,
white=white,
volume_template=mni152,
output_file=name,
labels=stat_labels,
interpolation=interpolation,
)
stat_volume = nib.load(name)
mask = (stat_volume.get_fdata() != 0) & (mni152.get_fdata() != 0)
stat_vector = stat_volume.get_fdata()[mask]
finally:
Path(name).unlink()
feature_names = []
correlations = np.zeros(len(feature_files))
logger.info("Running correlations with all Neurosynth features.")
for i in range(len(feature_files)):
feature_names.append(
re.search("__[A-Za-z0-9 ]+",
feature_files[i].stem)[0][2:]) # type: ignore
feature_data = nib.load(feature_files[i]).get_fdata()[mask]
keep = np.logical_not(
np.isnan(feature_data)
| np.isinf(feature_data)
| np.isnan(stat_vector)
| np.isinf(stat_vector))
correlations[i], _ = pearsonr(stat_vector[keep], feature_data[keep])
df = pd.DataFrame(correlations,
index=feature_names,
columns=["Pearson's r"])
return df.sort_values(by="Pearson's r", ascending=False)
def truncate_colormap(cmap, minval=0.0, maxval=1.0, n=100):
new_cmap = colors.LinearSegmentedColormap.from_list(
"trunc({n},{a:.2f},{b:.2f})".format(n=cmap.name, a=minval, b=maxval),
cmap(np.linspace(minval, maxval, n)),
)
return new_cmap
from nilearn import image, datasets
mean_img = image.load_img("../data/statistical_map.nii")
masker = input_data.NiftiMasker(mask_img=image.resample_to_img(
datasets.load_mni152_brain_mask(), mean_img, interpolation="nearest"))
# plot the image
masked_img = masker.fit_transform(mean_img)
unmasked_img = masker.inverse_transform(masked_img)
display = plotting.plot_glass_brain(
unmasked_img,
threshold=0,
cmap=plt.cm.viridis,
colorbar=False,
display_mode="xz",
axes=ax_glass_brain,
)
# ax_glass_brain.text(1, 1, '(a)', fontweight='bold')
ax1 = fig.add_axes([0.05, 0.12, 0.4, 0.03])
cmap = truncate_colormap(plt.cm.viridis, minval=0.5)
suj=89
meas=20
# Load Conditions file
eventname='F:/sim/event_sim.csv'
label=np.array(np.recfromcsv(eventname,names='s'))['s']
# Create session file
rest_mask = label == b'Rest'
# Prepare masker & Correct affine
template = load_mni152_template()
basc = datasets.fetch_atlas_basc_multiscale_2015(version='sym')['scale444']
orig_filename='F:/sim/template/restbaseline.nii.gz'
orig_img= image.load_img(orig_filename)
brainmask = load_mni152_brain_mask()
mem = Memory('nilearn_cache')
masker = NiftiLabelsMasker(labels_img = basc, mask_img = brainmask,
memory=mem, memory_level=1, verbose=0,
detrend=False, standardize=False,
high_pass=0.01,t_r=2,
resampling_target='labels')
masker.fit()
# Prep Classification
for s in np.arange(84,89):#range(suj):#
for n in range(meas):
sim_filename='F:/sim/sim_'+str(s+1)+'_'+ str(n+1)+ '.nii.gz'
if os.path.exists(sim_filename):
sim_img = image.load_img(sim_filename)
def parcellation(
sub_name, input_dir, out_dir, spatial_regularisation=10, n_clusters=200
):
"""
Will load a functional image, and parcellate the brain into different zones that tend to activate together
Parameters
----------
sub_name : string
name of the subject, used to access and save his data.
input_dir : string
path to where functional images can be loaded from
out_dir : string
path where files will be saved once computed
spatial_regularisation : float, default 10
once normalised, the spatial parameters a
n_clusters : int, default 200
The number of clusters to divide the image into
Notes
------
the "SUBJECTNAME_func_minimal.nii.gz" rsfMRI file is expected to be found in the input_dir,
in a folder named after the subject
no output, but saves three files in the out_dir, in the corresponding subject file
skward_regXX_parcellation.nii.gz : the ward parcellated file
ward_connectivity.npz : a sparse spatial adjacency matrix, voxel to voxel
ward_labels.npy
"""
raw_img = load_img(
os.path.join(input_dir, "{0}/{0}_func_minimal.nii.gz".format(sub_name)), 1
)
flat_img = (raw_img.get_fdata()).reshape(-1, raw_img.shape[3])
mask_img = load_mni152_brain_mask()
nifti_masker = NiftiMasker(
mask_img=mask_img, memory="nilearn_cache", memory_level=1
)
nifti_masker = nifti_masker.fit()
flat_img = nifti_masker.transform(raw_img)
plotting.plot_roi(mask_img, title="mask", display_mode="xz", cmap="prism")
mask = nifti_masker.mask_img_.get_fdata().astype(np.bool)
shape = mask.shape
connectivity = grid_to_graph(n_x=shape[0], n_y=shape[1], n_z=shape[2], mask=mask)
# weigh by coords
x_ = np.arange(61)
y_ = np.arange(73)
z_ = np.arange(61)
coords = np.array(np.meshgrid(x_, y_, z_, indexing="ij"))
assert np.all(coords[0, :, 0, 0] == x_)
assert np.all(coords[1, 0, :, 0] == y_)
assert np.all(coords[2, 0, 0, :] == z_)
coods_ni = new_img_like(raw_img, np.array(coords).T)
flat_coords = nifti_masker.transform(coods_ni)
# normalize
flat_img = (flat_img - flat_img.min()) / (flat_img.max() - flat_img.min())
flat_coords = (flat_coords - flat_coords.min()) / (
flat_coords.max() - flat_coords.min()
)
# add coords to data
input_data = np.vstack((flat_img, spatial_regularisation * flat_coords))
# Compute clustering
ward = AgglomerativeClustering(
n_clusters=n_clusters, linkage="ward", connectivity=connectivity
)
ward.fit(input_data.T)
labels = nifti_masker.inverse_transform(ward.labels_)
labels.to_filename(
os.path.join(
out_dir,
sub_name,
"skward_reg{}_parcellation.nii.gz".format(spatial_regularisation),
)
)
save_npz(
os.path.join(out_dir, sub_name, "ward_connectivity.npz",),
ward.connectivity.tocsr(),
)
np.save(os.path.join(out_dir, sub_name, "ward_labels.npy",), ward.labels_)
# CROSS MODAL CLASSIFICATION (cross-validated)
#n_p=0
#result_cv_tr_foot,permutation_scores_tr_foot, p_foot=permutation_score(pipeline,roi_foot_all,roi_hand_all,y_foot_all,groups,logo,n_p)
#print('Train FOOT - IMAG VS STIM',np.array(result_cv_tr_foot).mean(),p_foot)
#result_cv_tr_hand,permutation_scores_tr_hand, p_hand=permutation_score(pipeline,roi_hand_all,roi_foot_all,y_hand_all,groups,logo,n_p)
#print('Train HAND - IMAG VS STIM',np.array(result_cv_tr_hand).mean(),p_hand)
#result_cv_tr_imag,permutation_scores_tr_imag, p_imag=permutation_score(pipeline,roi_imag_all,roi_stim_all,y_imag_all,groups,logo,n_p)
#print('Train IMAG - HAND VS FOOT',np.array(result_cv_tr_imag).mean(),p_imag)
#result_cv_tr_stim,permutation_scores_tr_stim, p_stim=permutation_score(pipeline,roi_stim_all,roi_imag_all,y_stim_all,groups,logo,n_p)
#print('Train STIM - HAND VS FOOT',np.array(result_cv_tr_stim).mean(),p_stim)
# Prepare ploting
basc = datasets.fetch_atlas_basc_multiscale_2015(version='asym')['scale444']
brainmask = load_mni152_brain_mask()
masker = NiftiLabelsMasker(labels_img = basc, mask_img = brainmask,
memory_level=1, verbose=0,
detrend=True, standardize=False,
high_pass=0.01,t_r=2.28,
resampling_target='labels'
)
masker.fit()
pipeline.fit(roi_foot_all,y_foot_all)
coef_foot = pipeline.named_steps['svm'].coef_
weight_f = masker.inverse_transform(coef_foot)
plot_stat_map(weight_f, title='Train Imp',display_mode='z',cmap='bwr',threshold=0.4)
pipeline.fit(roi_hand_all,y_hand_all)
print("\nTop 10 neurosynth terms from downloaded images:\n")
for term_idx in np.argsort(total_scores)[-10:][::-1]:
print(vocabulary[term_idx])
######################################################################
# Reshape and mask images
# -----------------------
print("\nReshaping and masking images.\n")
with warnings.catch_warnings():
warnings.simplefilter('ignore', UserWarning)
warnings.simplefilter('ignore', DeprecationWarning)
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)
# Elaborate the ERP time windows
ERPs = {
'N1': [[1232, 1301], [0.101, 0.1348]],
'P2': [[1353, 1499], [0.160, 0.2314]],
'N2': [[1517, 1721], [0.2402, 0.3394]],
'P3': [[1722, 1927], [0.3403, 0.4404]],
'eLPP': [[1928, 2356], [0.4409, 0.6499]],
'lLPP': [[2357, 2868], [0.6504, 0.8999]]
}
# Import the eeg data
pos, time, fun = import_data(fname)
# Import the mni template
bm = load_mni152_brain_mask()
bm_data = bm.get_fdata()
Lx, Ly, Lz = bm_data.shape
# Apply shift and scaling to the data and reorder the axis in pos
shift = np.array([0, 40, 40])
scale = np.array([1, 1, 1])
ix, iy, iz = [1, 0, 2]
pos_ref = np.array([pos[:, ix], pos[:, iy], pos[:, iz]]).T
transformed_pos = pos_ref * scale[None, :] - shift[None, :]
# Plot a 2d crosscut overview to check that the scaling makes sense
plot_2D_crosscuts(bm_data, transformed_pos, fdir)
# Send the position to volume space
pos_volume = coord_transform(transformed_pos[:, 0], transformed_pos[:, 1],
def plot_4d_image_surface(components, colors, output_dir):
# Indices of the maps to outline
indices = []
mlab.options.offscreen = True
fig = mlab.figure(bgcolor=(1, 1, 1))
# Disable rendering to speed things up
fig.scene.disable_render = True
rng = np.random.RandomState(42)
mask = datasets.load_mni152_brain_mask().get_data() > 0
n_components = components.shape[-1]
threshold = np.percentile(np.abs(components.get_data()[mask]),
100. * (1 - 1. / n_components))
# To speed up when prototyping
# components = image.index_img(components, slice(0, 5))
actors = dict(left=[], right=[])
delayed = []
for i, (component, color) in enumerate(zip(
image.iter_img(components),
colors)):
if i in indices:
delayed.append((i, (component, color)))
else:
print('Component %i' % i)
component = image.math_img('np.abs(img)', img=component)
this_actors = add_surf_map(component, threshold=threshold,
color=tuple(color[:3]),
sides=['left', 'right'],
selected=i in indices,
inflate=1.001)
actors['left'].extend(this_actors['left'])
actors['right'].extend(this_actors['right'])
for i, (component, color) in delayed:
print('Component %i' % i)
component = image.math_img('np.abs(img)', img=component)
this_actors = add_surf_map(component, threshold=threshold,
color=tuple(color[:3]),
sides=['left', 'right'],
selected=i in indices,
inflate=1.001)
actors['left'].extend(this_actors['left'])
actors['right'].extend(this_actors['right'])
# Plot the background
for side in ['left', 'right']:
depth = surface.load_surf_data(fsaverage['sulc_%s' % side])
this_actors = plot_on_surf(-depth, sides=[side, ],
colormap='gray', inflate=.995)
actors[side].extend(this_actors[side])
# Enable rendering
fig.scene.disable_render = False
save_views(fig, 'components_3d', actors,
left_actors=actors['left'],
right_actors=actors['right'], output_dir=output_dir)
def surface_decode_nimare(
pial,
white,
stat_labels,
mask_labels,
interpolation="linear",
data_dir=None,
feature_group=None,
features=None,
):
"""Meta-analytic decoding of surface maps using NeuroSynth or Brainmap.
Parameters
----------
pial : str, BSPolyData, list
Path of a pial surface file, BSPolyData of a pial surface or a list
containing multiple of the aforementioned.
white : str, BSPolyData, list
Path of a white matter surface file, BSPolyData of a pial surface or a
list containing multiple of the aforementioned.
stat_labels : str, numpy.ndarray, list
Path to a label file for the surfaces, numpy array containing the
labels, or a list containing multiple of the aforementioned.
mask_labels : str, numpy.ndarray, list
Path to a mask file for the surfaces, numpy array containing the
mask, or a list containing multiple of the aforementioned. If None
all vertices are included in the mask. Defaults to None.
interpolation : str, optional
Either 'nearest' for nearest neighbor interpolation, or 'linear'
for trilinear interpolation, by default 'linear'.
data_dir : str, optional
The directory of the nimare dataset. If none exists, a new dataset will
be downloaded and saved to this path. If None, the directory defaults to
your home directory, by default None.
correction : str, optional
Multiple comparison correction. Valid options are None and 'fdr_bh',
by default 'fdr_bh'.
Returns
-------
pandas.DataFrame
Table with each label and the following values associated with each
label: ‘pForward’, ‘zForward’, ‘likelihoodForward’,‘pReverse’,
‘zReverse’, and ‘probReverse’.
"""
if data_dir is None:
data_dir = os.path.join(str(Path.home()), "nimare_data")
mni152 = load_mni152_brain_mask()
stat_image = tempfile.NamedTemporaryFile(suffix=".nii.gz")
mask_image = tempfile.NamedTemporaryFile(suffix=".nii.gz")
multi_surface_to_volume(
pial,
white,
mni152,
stat_labels,
stat_image.name,
interpolation=interpolation,
)
multi_surface_to_volume(
pial,
white,
mni152,
mask_labels,
mask_image.name,
interpolation="nearest",
)
dataset = fetch_nimare_dataset(data_dir, mask=mask_image.name, keep_neurosynth=True)
logging.info(
"If you use BrainStat's surface decoder, "
+ "please cite NiMARE (https://zenodo.org/record/4408504#.YBBPAZNKjzU))."
)
decoder = nimare.decode.continuous.CorrelationDecoder(
feature_group=feature_group, features=features
)
decoder.fit(dataset)
return decoder.transform(stat_image.name)
for term_idx in np.argsort(total_scores)[-10:][::-1]:
print(vocabulary[term_idx])
######################################################################
# Reshape and mask images
# -----------------------
print("\nReshaping and masking images.\n")
with warnings.catch_warnings():
warnings.simplefilter('ignore', UserWarning)
warnings.simplefilter('ignore', DeprecationWarning)
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:
import os.path as op
import numpy as np
import nibabel as nib
from nilearn import datasets, input_data
from nilearn.image import resample_to_img
from brainconn import utils
mask_img = datasets.load_mni152_brain_mask()
subjects = datasets.fetch_adhd(n_subjects=1)
power = datasets.fetch_coords_power_2011()
conf = subjects.confounds[0]
func_img = nib.load(subjects.func[0])
func_img = resample_to_img(func_img, mask_img)
coords = np.vstack((power.rois['x'], power.rois['y'], power.rois['z'])).T
spheres_masker = input_data.NiftiSpheresMasker(
seeds=coords, smoothing_fwhm=4, radius=5.,
detrend=True, standardize=True, low_pass=0.1, high_pass=0.01,
t_r=func_img.header.get_zooms()[-1])
timeseries = spheres_masker.fit_transform(func_img,
confounds=conf)
corr = np.corrcoef(timeseries.T)
np.savetxt(op.join(utils.get_resource_path(), 'example_corr.txt'), corr)
####################################################################
# First we load the ADHD200 data
from nilearn import datasets
import scipy as sp
from fmri_methods_sipi import rot_sub_data, hotelling_t2
import matplotlib.pyplot as plt
from nilearn import input_data
from nilearn.plotting import plot_roi, show, plot_stat_map
from nilearn.image.image import mean_img
from nilearn.image import index_img
from sklearn.decomposition import PCA
from scipy.stats import levene
from statsmodels.sandbox.stats.multicomp import multipletests
from nilearn.image import resample_to_img
std_msk = datasets.load_mni152_brain_mask()
gm_msk = datasets.fetch_icbm152_brain_gm_mask(threshold=0.3)
gm_msk.set_data_dtype(sp.float32)
affine_tar = 1.0 * sp.eye(4)
affine_tar[3, 3] = .3
adhd_dataset = datasets.fetch_adhd()
func_filenames = adhd_dataset.func # list of 4D nifti files for each subject
mean_func_img = mean_img(func_filenames[0])
gm_msk = resample_to_img(source_img=gm_msk,
target_img=mean_func_img,
interpolation='nearest')
#affine_tar,