def grouplevel_Ttest2(ListSubj,condition,method,mod,twin,clust_p):
wdir = "/neurospin/meg/meg_tmp/MTT_MEG_Baptiste/MEG"
# load a specific STC morphed on fsaverage to get shape info
stc0_path = (wdir + '/' + ListSubj[0] + '/mne_python/STCS/IcaCorr_' + mod +
'_' + ListSubj[0] + '_' + condition[0] + '_pick_oriNone_' +
method + '_ico-5-fwd-fsaverage.fif-rh.stc')
stc0 = mne.read_source_estimate(stc0_path)
ncond = len(condition)
nsub = len(ListSubj)
ntimes = len(stc0.times)
nvertices = stc0.data.shape[0]
# average individual STCs morphed on fsaverage for each cond
AVG_STC_cond = np.empty([nvertices, ntimes, nsub, ncond])
for s,subj in enumerate(ListSubj):
# multfact = 0
# for c,cond in enumerate(condition):
# stc_path = (wdir + '/' + subj + '/mne_python/STCS/IcaCorr_' + modality +
# '_' + subj + '_' + cond + '_pick_oriNone_' +
# method + '_ico-5-fwd-fsaverage.fif-rh.stc')
# stc = mne.read_source_estimate(stc_path)
# multfact = multfact + np.mean(np.mean(stc.data))
for c,cond in enumerate(condition):
stc_path = (wdir + '/' + subj + '/mne_python/STCS/IcaCorr_' + modality +
'_' + subj + '_' + cond + '_pick_oriNone_' +
method + '_ico-5-fwd-fsaverage.fif-rh.stc')
stc = mne.read_source_estimate(stc_path)
# AVG_STC_cond[:,:,s,c] = stc.data/(multfact/2)
AVG_STC_cond[:,:,s,c] = stc.data
# Compute statistic
np.random.seed(42)
# compute subject-by-subject mean difference
X = AVG_STC_cond[:,:,:,1] - AVG_STC_cond[:,:,:,0]
# smooth the data (optional)
#fsave_vertices = [np.arange(10242), np.arange(10242)]
#morph_mat = compute_morph_matrix('sample', 'fsaverage', sample_vertices,
# fsave_vertices, 20, subjects_dir)
#n_vertices_fsave = morph_mat.shape[0]
# optional: restrict computation to temporal window of interest
lower_bound = np.where(stc0.times >= twin[0])[0][0]
upper_bound = np.where(stc0.times >= twin[1])[0][0]
X = X[:,lower_bound:upper_bound,:]
con = mne.spatial_tris_connectivity(grade_to_tris(5))
X = np.transpose(X, [2, 1, 0])
p_threshold = clust_p
t_threshold = -stats.distributions.t.ppf(p_threshold / 2., nsub -1)
n_permutations = 1024
T_obs, clusters, cluster_p_values, H0 = mne.stats.spatio_temporal_cluster_1samp_test(X, connectivity=con, n_jobs=40,
threshold=t_threshold, n_permutations=n_permutations, verbose=True)
return T_obs, clusters, cluster_p_values, H0,stc0
def stat_clus(X, tstep, time_thre=0, n_per=8192, p_threshold=0.01, p=0.05, fn_clu_out=None):
print('Computing connectivity.')
connectivity = spatial_tris_connectivity(grade_to_tris(5))
# Note that X needs to be a multi-dimensional array of shape
# samples (subjects) x time x space, so we permute dimensions
X = np.transpose(X, [2, 1, 0])
n_subjects = X.shape[0]
fsave_vertices = [np.arange(X.shape[-1]/2), np.arange(X.shape[-1]/2)]
# Now let's actually do the clustering. This can take a long time...
# Here we set the threshold quite high to reduce computation.
t_threshold = -stats.distributions.t.ppf(p_threshold / 2., n_subjects - 1)
print('Clustering.')
max_step = int(time_thre * 0.001 / tstep) + 1
T_obs, clusters, cluster_p_values, H0 = clu = \
spatio_temporal_cluster_1samp_test(X, connectivity=connectivity, n_jobs=1, max_step=max_step,
threshold=t_threshold, n_permutations=n_per)
# Now select the clusters that are sig. at p < 0.05 (note that this value
# is multiple-comparisons corrected).
good_cluster_inds = np.where(cluster_p_values < p)[0]
print 'the amount of significant clusters are: %d' %good_cluster_inds.shape
###############################################################################
# Save the clusters as stc file
# ----------------------
assert good_cluster_inds.shape != 0, ('Current p_threshold is %f %p_thr,\
maybe you need to reset a lower p_threshold')
np.savez(fn_clu_out, clu=clu, tstep=tstep, fsave_vertices=fsave_vertices)
def run_permutation_ttest(tmin=None, tmax=None, p_threshold = 0.05, n_permutations=1024, inverse_method='dSPM', n_jobs=1):
for cond_id, cond_name in enumerate(events_id.keys()):
#todo: calc the 36
controls_data = get_morphed_epochs_stcs(tmin, tmax, cond_name, get_healthy_controls(),
36, inverse_method)
controls_data = abs(controls_data)
for patient in get_patients():
try:
print(patient, cond_name)
patient_data = get_morphed_epochs_stcs(tmin, tmax, cond_name, [patient], None, inverse_method)
patient_data = abs(patient_data)
print(patient_data.shape, controls_data.shape)
data = controls_data - patient_data
del patient_data
gc.collect()
data = np.transpose(data, [2, 1, 0])
connectivity = spatial_tris_connectivity(grade_to_tris(5))
t_threshold = -stats.distributions.t.ppf(p_threshold / 2., data.shape[0] - 1)
T_obs, clusters, cluster_p_values, H0 = \
spatio_temporal_cluster_1samp_test(data, connectivity=connectivity, n_jobs=n_jobs,
threshold=t_threshold, n_permutations=n_permutations)
results_file_name = op.join(LOCAL_ROOT_DIR, 'permutation_ttest_results', '{}_{}_{}'.format(patient, cond_name, inverse_method))
np.savez(results_file_name, T_obs=T_obs, clusters=clusters, cluster_p_values=cluster_p_values, H0=H0)
good_cluster_inds = np.where(cluster_p_values < 0.05)[0]
print('good_cluster_inds: {}'.format(good_cluster_inds))
except:
print('bummer! {}, {}'.format(patient, cond_name))
print(traceback.format_exc())
def per2test(X1, X2, p_thr, p, tstep, n_per=8192, fn_clu_out=None):
# Note that X needs to be a multi-dimensional array of shape
# samples (subjects) x time x space, so we permute dimensions
n_subjects1 = X1.shape[2]
n_subjects2 = X2.shape[2]
fsave_vertices = [np.arange(X1.shape[0]/2), np.arange(X1.shape[0]/2)]
X1 = np.transpose(X1, [2, 1, 0])
X2 = np.transpose(X2, [2, 1, 0])
X = [X1, X2]
# Now let's actually do the clustering. This can take a long time...
# Here we set the threshold quite high to reduce computation.
f_threshold = stats.distributions.f.ppf(1. - p_thr / 2., n_subjects1 - 1, n_subjects2 - 1)
print('Clustering.')
connectivity = spatial_tris_connectivity(grade_to_tris(5))
T_obs, clusters, cluster_p_values, H0 = clu = \
spatio_temporal_cluster_test(X, n_permutations=n_per,
connectivity=connectivity, n_jobs=1,
threshold=f_threshold)
# Now select the clusters that are sig. at p < 0.05 (note that this value
# is multiple-comparisons corrected).
good_cluster_inds = np.where(cluster_p_values < p)[0]
print 'the amount of significant clusters are: %d' %good_cluster_inds.shape
###############################################################################
# Save the clusters as stc file
# ----------------------
assert good_cluster_inds.shape != 0, ('Current p_threshold is %f %p_thr,\
maybe you need to reset a lower p_threshold')
np.savez(fn_clu_out, clu=clu, tstep=tstep, fsave_vertices=fsave_vertices)
def grouplevel_spatial_linreg(ListSubj,condition,method,mod,twin,clust_p):
wdir = "/neurospin/meg/meg_tmp/MTT_MEG_Baptiste/MEG"
# load a specific STC morphed on fsaverage to get shape info
stc0_path = (wdir + '/' + ListSubj[0] + '/mne_python/STCS/IcaCorr_' + mod +
'_' + ListSubj[0] + '_' + condition[0] + '_pick_oriNone_' +
method + '_ico-5-fwd-fsaverage.fif-rh.stc')
stc0 = mne.read_source_estimate(stc0_path)
ncond = len(condition)
nsub = len(ListSubj)
ntimes = len(stc0.times)
nvertices = stc0.data.shape[0]
# average individual STCs morphed on fsaverage for each cond
AVG_STC_cond = np.empty([nvertices, ntimes, nsub, ncond])
for s,subj in enumerate(ListSubj):
for c,cond in enumerate(condition):
stc_path = (wdir + '/' + subj + '/mne_python/STCS/IcaCorr_' + modality +
'_' + subj + '_' + cond + '_pick_oriNone_' +
method + '_ico-5-fwd-fsaverage.fif-rh.stc')
stc = mne.read_source_estimate(stc_path)
AVG_STC_cond[:,:,s,c] = stc.data
# optional: restrict computation to temporal window of interest
lower_bound = np.where(stc0.times >= twin[0])[0][0]
upper_bound = np.where(stc0.times >= twin[1])[0][0]
con = mne.spatial_tris_connectivity(grade_to_tris(5))
# array of shapes (obs, time, vertices)
X, Xpred = [], []
for c,cond in enumerate(condition):
X.append(np.mean(np.transpose(AVG_STC_cond[:,lower_bound: upper_bound,:,c], [2, 1, 0]),1))
Xpred.append(np.ones((nsub,nvertices))*(c+1))
X = np.array(X)
Xpred = np.array(Xpred)
Reg = np.zeros((nsub,nvertices))
# lienar regression per vertex
for s in range(nsub):
for v in range(nvertices):
tmp = stats.linregress(X[:,s,v],Xpred[:,s,v])
Reg[s,v] = tmp.slope
if clust_p == None:
T_obs, clu, clu_p_val, H0 = mne.stats.permutation_cluster_1samp_test(Reg, None,
connectivity=con, n_jobs=8,
verbose=True, seed = 666)
else:
p_threshold = clust_p
t_threshold = -stats.distributions.t.ppf(p_threshold / 2., nsub -1)
T_obs, clu, clu_p_val, H0 = mne.stats.permutation_cluster_1samp_test(Reg, threshold=t_threshold,
connectivity=con, n_jobs=8,
verbose=True, seed = 666)
return T_obs, clu, clu_p_val, H0,stc0
def grouplevel_stats(ListSubj,condition,method,mod,twin,clust_p):
wdir = "/neurospin/meg/meg_tmp/MTT_MEG_Baptiste/MEG"
# load a specific STC morphed on fsaverage to get shape info
stc0_path = (wdir + '/' + ListSubj[0] + '/mne_python/STCS/IcaCorr_' + mod +
'_' + ListSubj[0] + '_' + condition[0] + '_pick_oriNone_' +
method + '_ico-5-fwd-fsaverage.fif-rh.stc')
stc0 = mne.read_source_estimate(stc0_path)
stc0.crop(-0.2,5)
ncond = len(condition)
nsub = len(ListSubj)
ntimes = len(stc0.times)
nvertices = stc0.data.shape[0]
# average individual STCs morphed on fsaverage for each cond
AVG_STC_cond = np.empty([nvertices, ntimes, nsub, ncond])
for s,subj in enumerate(ListSubj):
for c,cond in enumerate(condition):
stc_path = (wdir + '/' + subj + '/mne_python/STCS/IcaCorr_' + modality +
'_' + subj + '_' + cond + '_pick_oriNone_' +
method + '_ico-5-fwd-fsaverage.fif-rh.stc')
stc = mne.read_source_estimate(stc_path)
stc0.crop(-0.2,5)
AVG_STC_cond[:,:,s,c] = stc.data
# optional: restrict computation to temporal window of interest
lower_bound = np.where(stc0.times >= twin[0])[0][0]
upper_bound = np.where(stc0.times >= twin[1])[0][0]
con = mne.spatial_tris_connectivity(grade_to_tris(5))
# array of shapes (obs, time, vertices)
X = []
for c,cond in enumerate(condition):
X.append(np.transpose(AVG_STC_cond[:,lower_bound: upper_bound,:,c], [2, 1, 0]))
effects = 'A'
factor_levels = [3]
# get f-values only.
def mystat_fun(*args):
return f_mway_rm(np.swapaxes(args, 1, 0), factor_levels=factor_levels,
effects=effects, return_pvals=False)[0]
p_threshold = clust_p
f_threshold = mne.stats.f_threshold_mway_rm(nsub,factor_levels = factor_levels, effects = effects,pvalue= p_threshold)
T_obs, clu, clu_p_val, H0 = mne.stats.spatio_temporal_cluster_test(X, threshold=f_threshold,stat_fun=mystat_fun,
connectivity=con, n_jobs=1,
verbose=True, seed = 666)
return T_obs, clu, clu_p_val, H0,stc0
def per2test(X1, X2, p_thr, p, tstep, n_per=8192, fn_clu_out=None):
'''
Calculate significant clusters using 2 sample ftest.
Parameter
---------
X1, X2: array
The shape of X should be (Vertices, timepoints, subjects)
tstep: float
The interval between timepoints.
n_per: int
The permutation for ttest.
p_thr: float
The significant p_values.
p: float
The corrected p_values for comparisons.
fn_clu_out: string
The fnname for saving clusters.
'''
# Note that X needs to be a multi-dimensional array of shape
# samples (subjects) x time x space, so we permute dimensions
n_subjects1 = X1.shape[2]
n_subjects2 = X2.shape[2]
fsave_vertices = [np.arange(X1.shape[0]/2), np.arange(X1.shape[0]/2)]
X1 = np.transpose(X1, [2, 1, 0])
X2 = np.transpose(X2, [2, 1, 0])
X = [X1, X2]
# Now let's actually do the clustering. This can take a long time...
# Here we set the threshold quite high to reduce computation.
f_threshold = stats.distributions.f.ppf(1. - p_thr / 2., n_subjects1 - 1,
n_subjects2 - 1)
# t_threshold = stats.distributions.t.ppf(1. - p_thr / 2., n_subjects1 - 1,
# n_subjects2 - 1)
print('Clustering...')
connectivity = spatial_tris_connectivity(grade_to_tris(5))
T_obs, clusters, cluster_p_values, H0 = clu = \
spatio_temporal_cluster_test(X, n_permutations=n_per, #step_down_p=0.001,
connectivity=connectivity, n_jobs=1,
# threshold=t_threshold, stat_fun=stats.ttest_ind)
threshold=f_threshold)
# Now select the clusters that are sig. at p < 0.05 (note that this value
# is multiple-comparisons corrected).
good_cluster_inds = np.where(cluster_p_values < p)[0]
print 'the amount of significant clusters are: %d' % good_cluster_inds.shape
# Save the clusters as stc file
np.savez(fn_clu_out, clu=clu, tstep=tstep, fsave_vertices=fsave_vertices)
assert good_cluster_inds.shape != 0, ('Current p_threshold is %f %p_thr,\
maybe you need to reset a lower p_threshold')
def perform_statistics_2(morphed_data, parameter_cache, vector, p_value=None):
"""Performs the statistical analysis using spatial_tris_connectivity.
:param morphed_data: Morphed data obtained from morph_data function
:param parameter_cache: Morphed parameter cache obtained from morph_data function.
:param vector: Method to perform modelling ('sLORETA' etc.)
:param p_value: Statistical p-value
:return: clu, good_cluster_inds
"""
# Unpack parameter cache dictionary
n_subjects = parameter_cache['n_subjects']
n_times = parameter_cache['n_times']
# Take on the absolute
X = np.abs(morphed_data)
# Obtain the paired contrast
if vector is False:
X = X[:, :, :, 0] - X[:, :, :,
1] # Dimension is (space, time, subjects)
else:
X = X[:, :, :, :,
0] - X[:, :, :, :,
1] # Dimension is (space, vector, time, subjects)
print('Computing connectivity... ')
connectivity_2 = mne.spatio_temporal_tris_connectivity(
grade_to_tris(5), n_times)
# Note that X needs to be a multi-dimensional array of shape [samples (subjects) x time x space]
if vector is False:
X = np.transpose(X, [2, 1, 0])
else:
X = np.transpose(X, [3, 2, 1, 0]) ##### TO DOUBLE CHECK #####
# Perform the clustering
p_threshold = p_value # 0.001
t_threshold = -stats.distributions.t.ppf(p_threshold / 2., n_subjects - 1)
print('Clustering... ')
T_obs, clusters, cluster_p_values, H0 = spatio_temporal_cluster_1samp_test(
X, connectivity=connectivity_2, n_jobs=1, threshold=t_threshold)
# Pack the outputs into tuple
clu = (T_obs, clusters, cluster_p_values, H0)
# Select the clusters that are sig. at p < p_value (Note this value is multiple-comparisons corrected)
good_cluster_inds = np.where(cluster_p_values < p_value)[0]
return clu, good_cluster_inds
def permutation_test_on_source_data_with_spatio_temporal_clustering(controls_data, patient_data, patient, cond_name,
tstep, n_permutations, inverse_method='dSPM', n_jobs=6):
try:
print('permutation_test: patient {}, cond {}'.format(patient, cond_name))
connectivity = spatial_tris_connectivity(grade_to_tris(5))
# Note that X needs to be a list of multi-dimensional array of shape
# samples (subjects_k) x time x space, so we permute dimensions
print(controls_data.shape, patient_data.shape)
X = [controls_data, patient_data]
p_threshold = 0.05
f_threshold = stats.distributions.f.ppf(1. - p_threshold / 2.,
controls_data.shape[0] - 1, 1)
print('Clustering. thtreshold = {}'.format(f_threshold))
T_obs, clusters, cluster_p_values, H0 = clu =\
spatio_temporal_cluster_test(X, connectivity=connectivity, n_jobs=n_jobs, threshold=10, n_permutations=n_permutations)
results_file_name = op.join(LOCAL_ROOT_DIR, 'clusters_results', '{}_{}_{}'.format(patient, cond_name, inverse_method))
np.savez(results_file_name, T_obs=T_obs, clusters=clusters, cluster_p_values=cluster_p_values, H0=H0)
# Now select the clusters that are sig. at p < 0.05 (note that this value
# is multiple-comparisons corrected).
good_cluster_inds = np.where(cluster_p_values < 0.05)[0]
###############################################################################
# Visualize the clusters
print('Visualizing clusters.')
# Now let's build a convenient representation of each cluster, where each
# cluster becomes a "time point" in the SourceEstimate
fsave_vertices = [np.arange(10242), np.arange(10242)]
stc_all_cluster_vis = summarize_clusters_stc(clu, tstep=tstep,
vertices=fsave_vertices,
subject='fsaverage')
stc_all_cluster_vis.save(op.join(LOCAL_ROOT_DIR, 'stc_clusters', '{}_{}_{}'.format(patient, cond_name, inverse_method)), ftype='h5')
# # Let's actually plot the first "time point" in the SourceEstimate, which
# # shows all the clusters, weighted by duration
# # blue blobs are for condition A != condition B
# brain = stc_all_cluster_vis.plot('fsaverage', 'inflated', 'both',
# subjects_dir=subjects_dir, clim='auto',
# time_label='Duration significant (ms)')
# brain.set_data_time_index(0)
# brain.show_view('lateral')
# brain.save_image('clusters.png')
except:
print('bummer! {}, {}'.format(patient, cond_name))
print(traceback.format_exc())
def stat_clus(X, tstep, n_per=8192, p_threshold=0.01, p=0.05, fn_clu_out=None):
'''
Calculate significant clusters using 1sample ttest.
Parameter
---------
X: array
The shape of X should be (Vertices, timepoints, subjects)
tstep: float
The interval between timepoints.
n_per: int
The permutation for ttest.
p_threshold: float
The significant p_values.
p: float
The corrected p_values for comparisons.
fn_clu_out: string
The fnname for saving clusters.
'''
print('Computing connectivity.')
connectivity = spatial_tris_connectivity(grade_to_tris(5))
# Note that X needs to be a multi-dimensional array of shape
# samples (subjects) x time x space, so we permute dimensions
X = np.transpose(X, [2, 1, 0])
n_subjects = X.shape[0]
fsave_vertices = [np.arange(X.shape[-1]/2), np.arange(X.shape[-1]/2)]
# Now let's actually do the clustering. This can take a long time...
# Here we set the threshold quite high to reduce computation.
t_threshold = -stats.distributions.t.ppf(p_threshold / 2., n_subjects - 1)
print('Clustering.')
T_obs, clusters, cluster_p_values, H0 = clu = \
spatio_temporal_cluster_1samp_test(X, connectivity=connectivity,
n_jobs=1, threshold=t_threshold,
n_permutations=n_per)
# Now select the clusters that are sig. at p < 0.05 (note that this value
# is multiple-comparisons corrected).
good_cluster_inds = np.where(cluster_p_values < p)[0]
print 'the amount of significant clusters are: %d' %good_cluster_inds.shape
# Save the clusters as stc file
np.savez(fn_clu_out, clu=clu, tstep=tstep, fsave_vertices=fsave_vertices)
assert good_cluster_inds.shape != 0, ('Current p_threshold is %f %p_thr,\
maybe you need to reset a lower p_threshold')
def sublevel_spatial_stats(X, subject, index, stc, clust_p, modality,
method, condition):
con = mne.spatial_tris_connectivity(grade_to_tris(5))
# morphing data
subjects_dir = '/neurospin/meg/meg_tmp/MTT_MEG_Baptiste/mri/'
fsave_vertices = [np.arange(10242), np.arange(10242)]
subject_vertices = [stc.lh_vertno, stc.rh_vertno]
morph_mat = compute_morph_matrix(subject, 'fsaverage',
subject_vertices, fsave_vertices, 20,
subjects_dir)
n_vertices_fsave = morph_mat.shape[0]
nobs = index[0]
ncond = len(index)
ntimes = X.shape[2]
nvertices = stc.lh_vertno.shape[0] + stc.rh_vertno.shape[0]
# We have to change the shape for the dot() to work properly
X = np.transpose(X, [1, 2, 0]) # vertices * times * obs
X = X.reshape(nvertices, ntimes * nobs * ncond)
print('Morphing data.')
X = morph_mat.dot(X) # morph_mat is a sparse matrix
X = X.reshape(n_vertices_fsave, ntimes, nobs, ncond)
# list of array of shapes ncond*(obs, time, vertices)
X = np.transpose(X, [2, 1, 0, 3])
X = [np.squeeze(x) for x in np.split(X, 3, axis=-1)]
# average over time to clusterize only on spatial dimension
X = [np.mean(x, 1) for x in X]
p_threshold = clust_p
f_threshold = stats.distributions.f.ppf(1 - p_threshold, ncond - 1,
ncond * (nobs - 1))
# use default function one-way anova (not repeated measures!)
F_obs, clu, clu_p_val, H0 = mne.stats.permutation_cluster_test(
X,
threshold=f_threshold,
connectivity=con,
n_jobs=4,
verbose=True,
seed=666)
wdir = "/neurospin/meg/meg_tmp/MTT_MEG_Baptiste/MEG/"
save_path = (wdir + subject + '/mne_python/STATS')
if not os.path.exists(save_path):
os.makedirs(save_path)
# save cluster stats
spatial_clust_F = np.array((F_obs, clu, clu_p_val, H0))
np.save((save_path + '/' + modality + '_' + 'cluster_stats_' +
"_vs_".join(condition)), spatial_clust_F)
# save F-Map
tmp = F_obs
tmp = tmp[:, np.newaxis]
fsave_vertices = [np.arange(10242), np.arange(10242)]
stc_Ftest = mne.SourceEstimate(tmp, fsave_vertices, 0, stc.tstep)
stc_Ftest.save(
(save_path + '/' + modality + '_' + "_vs_".join(condition)))
# to create probability maps: threshold the map at p < 0.05 and binarize
ind = np.where(F_obs >= f_threshold)[0]
VertKept = np.empty((len(F_obs)))
for v, vertices in enumerate(VertKept):
if v in ind:
VertKept[v] = 1
else:
VertKept[v] = 0
VertKept = VertKept[:, np.newaxis]
fsave_vertices = [np.arange(10242), np.arange(10242)]
stc_Ftest = mne.SourceEstimate(VertKept, fsave_vertices, 0, stc.tstep)
stc_Ftest.save((save_path + '/' + modality + '_' + 'BinForProb_' +
"_vs_".join(condition)))
return F_obs, clu, clu_p_val, H0
def sample2_clus(fn_list,
n_per=8192,
pthr=0.01,
p=0.05,
tail=0,
del_vers=None,
n_jobs=1):
'''
Calculate significant clusters using 2 sample ftest.
Parameter
---------
fn_list: list
Paths of group arrays
n_per: int
The permutation for ttest.
pct: int or float.
The percentile of the baseline distribution.
p: float
The corrected p_values for comparisons.
del_vers: None or _exclu_vers
If is '_exclu_vers', delete the vertices in the medial wall.
'''
for fn_npz in fn_list:
fn_path = os.path.dirname(fn_npz)
name = os.path.basename(fn_npz)
#fn_out = fn_path + '/clu2sample_%s' %name[:name.rfind('.npz')] + '_%d_pct%.2f.npz' %(n_per, pct)
fn_out = fn_path + '/clu2sample_%s' % name[:name.rfind(
'.npz')] + '_%d_%dtail_pthr%.4f.npz' % (n_per, 1 +
(tail == 0), pthr)
npz = np.load(fn_npz)
tstep = npz['tstep'].flatten()[0]
# Note that X needs to be a multi-dimensional array of shape
# samples (subjects) x time x space, so we permute dimensions
X = npz['X']
ppf = stats.f.ppf
tail = 1 # tail = we are interested in an increase of variance only
p_thresh = pthr / (
1 + (tail == 0)
) # we can also adapt this to p=0.01 if the cluster size is too large
n_samples_per_group = [len(x) for x in X]
f_threshold = ppf(1. - p_thresh, *n_samples_per_group)
if np.sign(tail) < 0:
f_threshold = -f_threshold
fsave_vertices = [
np.arange(X.shape[-1] / 2),
np.arange(X.shape[-1] / 2)
]
print('Clustering...')
connectivity = spatial_tris_connectivity(grade_to_tris(5))
T_obs, clusters, cluster_p_values, H0 = clu = \
spatio_temporal_cluster_test(X, n_permutations=n_per, #step_down_p=0.001,
connectivity=connectivity, n_jobs=n_jobs,
# threshold=t_threshold, stat_fun=stats.ttest_ind)
threshold=f_threshold, spatial_exclude=del_vers, tail=tail)
# Now select the clusters that are sig. at p < 0.05 (note that this value
# is multiple-comparisons corrected).
good_cluster_inds = np.where(cluster_p_values < p)[0]
print 'the amount of significant clusters are: %d' % good_cluster_inds.shape
# Save the clusters as stc file
np.savez(fn_out, clu=clu, tstep=tstep, fsave_vertices=fsave_vertices)
assert good_cluster_inds.shape != 0, (
'Current p_threshold is %f %p_thr,\
maybe you need to reset a lower p_threshold'
)
for fname in files:
# exclude due to psychotropic factors
if fname.find('WCEYBWMO') < 0:
stc = mne.read_source_estimate(fname[:-7])
if after_zero:
stc.crop(tmin=0)
X.append(stc.data)
subj = fname.split('/')[-1].split('_')[0]
gf_loc.append(np.nonzero(gf.maskid == subj)[0][0])
if len(X) != len(gf_loc):
print 'Mismatch between subject data and gf!'
# re-sort subject order in gf to match X
gf = gf.iloc[gf_loc]
connectivity = mne.spatial_tris_connectivity(mne.grade_to_tris(5))
for thresh in p_threshold:
if my_test == 'nvVSper':
g0 = [X[i].T for i, group in enumerate(gf.group) if group == 'NV']
g1 = [X[i].T for i, group in enumerate(gf.group) if group == 'persistent']
data = [np.array(g0), np.array(g1)]
stat_obs, clusters, p_values, H0 = \
mne.stats.spatio_temporal_cluster_test(data, n_jobs=njobs,
threshold=thresh,
connectivity=connectivity,
stat_fun=group_comparison,
tail=1,
n_permutations=nperms,
verbose=True)
elif my_test == 'nvVSrem':
return f_twoway_rm(data,
factor_levels=factor_levels,
effects=effects,
return_pvals=return_pvals)[0]
# drop p-values (empty array).
# Note. for further details on this ANOVA function consider the
# corresponding time frequency example.
###############################################################################
# Compute clustering statistic
# To use an algorithm optimized for spatio-temporal clustering, we
# just pass the spatial connectivity matrix (instead of spatio-temporal)
source_space = grade_to_tris(5)
# as we only have one hemisphere we need only need half the connectivity
lh_source_space = source_space[source_space[:, 0] < 10242]
print('Computing connectivity.')
connectivity = spatial_tris_connectivity(lh_source_space)
# Now let's actually do the clustering. Please relax, on a small
# notebook and one single thread only this will take a couple of minutes ...
pthresh = 0.0005
f_thresh = f_threshold_twoway_rm(n_subjects, factor_levels, effects, pthresh)
# To speed things up a bit we will ...
n_permutations = 100 # ... run fewer permutations (reduces sensitivity)
print('Clustering.')
T_obs, clusters, cluster_p_values, H0 = clu = \
def apply_sigSTC(fn_list_v, vevent, mevent, method='dSPM', vtmin=0., vtmax=0.35,
mtmin=-0.3, mtmax=0.05, radius=10.0):
from mne import spatial_tris_connectivity, grade_to_tris
from mne.stats import spatio_temporal_cluster_test
from scipy import stats as stats
X1, X2 = [], []
stcs_trial = []
for fn_v in fn_list_v:
fn_m = fn_v[: fn_v.rfind('evtW')] + 'evtW_%s_bc_norm_1-lh.stc' %mevent
stc_v = mne.read_source_estimate(fn_v)
stcs_trial.append(stc_v.copy())
stc_m = mne.read_source_estimate(fn_m)
stc_v.resample(200)
stc_m.resample(200)
X1.append(stc_v.copy().crop(vtmin, vtmax).data)
X2.append(stc_m.copy().crop(mtmin, mtmax).data)
stcs_path = subjects_dir+'/fsaverage/%s_ROIs/conditions/' %method
reset_directory(stcs_path)
fn_avg = stcs_path + '%s' %(vevent)
stcs = np.array(stcs_trial)
stc_avg = np.sum(stcs, axis=0)/stcs.shape[0]
stc_avg.save(fn_avg, ftype='stc')
X1 = np.array(X1).transpose(0, 2, 1)
X2 = np.array(X2).transpose(0, 2, 1)
###############################################################################
# Compute statistic
# To use an algorithm optimized for spatio-temporal clustering, we
# just pass the spatial connectivity matrix (instead of spatio-temporal)
print('Computing connectivity.')
connectivity = spatial_tris_connectivity(grade_to_tris(5))
# Note that X needs to be a list of multi-dimensional array of shape
# samples (subjects_k) x time x space, so we permute dimensions
X = [X1, X2]
# Now let's actually do the clustering. This can take a long time...
# Here we set the threshold quite high to reduce computation.
p_threshold = 0.0001
f_threshold = stats.distributions.f.ppf(1. - p_threshold / 2.,
X1.shape[0] - 1, X1.shape[0] - 1)
print('Clustering.')
clu = spatio_temporal_cluster_test(X, connectivity=connectivity, n_jobs=2,
threshold=f_threshold)
# Now select the clusters that are sig. at p < 0.05 (note that this value
# is multiple-comparisons corrected).
#fsave_vertices = [np.arange(10242), np.arange(10242)]
tstep = stc_v.tstep
#stc_all_cluster_vis = summarize_clusters_stc(clu, tstep=tstep,
# vertices=fsave_vertices,
# subject='fsaverage')
#stc_sig = stc_all_cluster_vis.mean()
#fn_sig = subjects_dir+'/fsaverage/%s_ROIs/%s' %(method,vevent)
#stc_sig.save(fn_sig)
tstep = stc_v.tstep
T_obs, clusters, clu_pvals, _ = clu
n_times, n_vertices = T_obs.shape
good_cluster_inds = np.where(clu_pvals < 0.05)[0]
seeds = []
# Build a convenient representation of each cluster, where each
# cluster becomes a "time point" in the SourceEstimate
T_obs = abs(T_obs)
if len(good_cluster_inds) > 0:
data = np.zeros((n_vertices, n_times))
for cluster_ind in good_cluster_inds:
data.fill(0)
v_inds = clusters[cluster_ind][1]
t_inds = clusters[cluster_ind][0]
data[v_inds, t_inds] = T_obs[t_inds, v_inds]
# Store a nice visualization of the cluster by summing across time
data = np.sign(data) * np.logical_not(data == 0) * tstep
seed = np.argmax(data.sum(axis=-1))
seeds.append(seed)
min_subject = 'fsaverage'
labels_path = subjects_dir + '/fsaverage/dSPM_ROIs/%s/ini' %vevent
reset_directory(labels_path)
seeds = np.array(seeds)
non_index_lh = seeds[seeds < 10242]
if non_index_lh.shape != []:
func_labels_lh = mne.grow_labels(min_subject, non_index_lh,
extents=radius, hemis=0,
subjects_dir=subjects_dir, n_jobs=1)
i = 0
while i < len(func_labels_lh):
func_label = func_labels_lh[i]
func_label.save(labels_path + '/%s_%d' %(vevent, i))
i = i + 1
seeds_rh = seeds - 10242
non_index_rh = seeds_rh[seeds_rh > 0]
if non_index_rh.shape != []:
func_labels_rh = mne.grow_labels(min_subject, non_index_rh,
extents=radius, hemis=1,
subjects_dir=subjects_dir, n_jobs=1)
# right hemisphere definition
j = 0
while j < len(func_labels_rh):
func_label = func_labels_rh[j]
func_label.save(labels_path + '/%s_%d' %(vevent, j))
j = j + 1
def sample1_clus(fn_list,
n_per=8192,
pct=99,
p=0.01,
tail=1,
del_vers=None,
n_jobs=1):
'''
Calculate significant clusters using 1sample ttest.
Parameter
---------
fn_list: list
Paths of group arrays
n_per: int
The permutation for ttest.
pct: int or float.
The percentile of the baseline distribution.
p: float
The corrected p_values for comparisons.
tail: 1 or 0
if tail=1, that is 1 tail test
if tail=0, that is 2 tail test
del_vers: None or _exclu_vers
If is '_exclu_vers', delete the vertices in the medial wall.
'''
print('Computing connectivity.')
connectivity = spatial_tris_connectivity(grade_to_tris(5))
# Using the percentile of baseline array as the distribution threshold
for fn_npz in fn_list:
npz = np.load(fn_npz)
tstep = npz['tstep'].flatten()[0]
# Note that X needs to be a multi-dimensional array of shape
# samples (subjects) x time x space, so we permute dimensions
X = npz['X']
X_b = X[1]
X = X[0]
fn_path = os.path.dirname(fn_npz)
name = os.path.basename(fn_npz)
if tail == 1:
fn_out = fn_path + '/clu1sample_%s' % name[:name.rfind(
'.npz')] + '_%d_%dtail_pct%.3f.npz' % (n_per, tail, pct)
X = np.abs(X)
t_threshold = np.percentile(np.abs(X_b), pct)
elif tail == 0:
fn_out = fn_path + '/clu1sample_%s' % name[:name.rfind(
'.npz')] + '_%d_%dtail_pct%.3f.npz' % (n_per, tail + 2, pct)
t_threshold = np.percentile(X_b, pct)
fsave_vertices = [
np.arange(X.shape[-1] / 2),
np.arange(X.shape[-1] / 2)
]
#n_subjects = X.shape[0]
#t_threshold = -stats.distributions.t.ppf(0.01/(1+(tail==0)), n_subjects-1)
print('Clustering.')
T_obs, clusters, cluster_p_values, H0 = clu = \
spatio_temporal_cluster_1samp_test(X, connectivity=connectivity,
n_jobs=n_jobs, threshold=t_threshold,
n_permutations=n_per, tail=tail, spatial_exclude=del_vers)
# Now select the clusters that are sig. at p < 0.05 (note that this value
# is multiple-comparisons corrected).
good_cluster_inds = np.where(cluster_p_values < p)[0]
print 'the amount of significant clusters are: %d' % good_cluster_inds.shape
# Save the clusters as stc file
np.savez(fn_out, clu=clu, tstep=tstep, fsave_vertices=fsave_vertices)
assert good_cluster_inds.shape != 0, (
'Current p_threshold is %f %p_thr,\
maybe you need to reset a lower p_threshold'
)
# subjects X conditions X observations (optional).
# The following expression catches the list input
# and swaps the first and the second dimension, and finally calls ANOVA.
return f_twoway_rm(np.swapaxes(args, 1, 0), factor_levels=factor_levels,
effects=effects, return_pvals=return_pvals)[0]
# get f-values only.
# Note. for further details on this ANOVA function consider the
# corresponding time frequency example.
###############################################################################
# Compute clustering statistic
# To use an algorithm optimized for spatio-temporal clustering, we
# just pass the spatial connectivity matrix (instead of spatio-temporal)
source_space = grade_to_tris(5)
# as we only have one hemisphere we need only need half the connectivity
lh_source_space = source_space[source_space[:, 0] < 10242]
print('Computing connectivity.')
connectivity = spatial_tris_connectivity(lh_source_space)
# Now let's actually do the clustering. Please relax, on a small
# notebook and one single thread only this will take a couple of minutes ...
pthresh = 0.0005
f_thresh = f_threshold_twoway_rm(n_subjects, factor_levels, effects, pthresh)
# To speed things up a bit we will ...
n_permutations = 128 # ... run fewer permutations (reduces sensitivity)
print('Clustering.')
T_obs, clusters, cluster_p_values, H0 = clu = \
X2 = np.random.randn(n_vertices_fsave, n_times, n_subjects2) * 10
X1[:, :, :] += stc.data[:, :, np.newaxis]
# make the activity bigger for the second set of subjects
X2[:, :, :] += 3 * stc.data[:, :, np.newaxis]
# We want to compare the overall activity levels for each subject
X1 = np.abs(X1) # only magnitude
X2 = np.abs(X2) # only magnitude
###############################################################################
# Compute statistic
# To use an algorithm optimized for spatio-temporal clustering, we
# just pass the spatial connectivity matrix (instead of spatio-temporal)
print('Computing connectivity.')
connectivity = spatial_tris_connectivity(grade_to_tris(5))
# Note that X needs to be a list of multi-dimensional array of shape
# samples (subjects_k) x time x space, so we permute dimensions
X1 = np.transpose(X1, [2, 1, 0])
X2 = np.transpose(X2, [2, 1, 0])
X = [X1, X2]
# Now let's actually do the clustering. This can take a long time...
# Here we set the threshold quite high to reduce computation.
p_threshold = 0.0001
f_threshold = stats.distributions.f.ppf(1. - p_threshold / 2.,
n_subjects1 - 1, n_subjects2 - 1)
print('Clustering.')
T_obs, clusters, cluster_p_values, H0 = clu =\
spatio_temporal_cluster_test(X, connectivity=connectivity, n_jobs=2,
for fname in files:
# exclude due to psychotropic factors
if fname.find('WCEYBWMO') < 0:
stc = mne.read_source_estimate(fname[:-7])
if after_zero:
stc.crop(tmin=0)
X.append(stc.data)
subj = fname.split('/')[-1].split('_')[0]
gf_loc.append(np.nonzero(gf.maskid == subj)[0][0])
if len(X) != len(gf_loc):
print 'Mismatch between subject data and gf!'
# re-sort subject order in gf to match X
gf = gf.iloc[gf_loc]
connectivity = mne.spatial_tris_connectivity(mne.grade_to_tris(5))
for thresh in p_threshold:
if my_test == 'nvVSper':
g0 = [X[i].T for i, group in enumerate(gf.group) if group == 'NV']
g1 = [
X[i].T for i, group in enumerate(gf.group) if group == 'persistent'
]
data = [np.array(g0), np.array(g1)]
stat_obs, clusters, p_values, H0 = \
mne.stats.spatio_temporal_cluster_test(data, n_jobs=njobs,
threshold=thresh,
connectivity=connectivity,
stat_fun=group_comparison,
tail=1,
n_permutations=nperms,
for f in np.arange(1, 2):
print('Clustering: ', f_label[f], '/ k:', k, '/ w: ', w)
X_SD = np.zeros([18, 20484])
X_LD = np.zeros([18, 20484])
for i in np.arange(0, 18):
X_SD[i, :] = (abs(
my_stc_coh_SD[i][w][k][f].data)).reshape(20484)
X_LD[i, :] = (abs(
my_stc_coh_LD[i][w][k][f].data)).reshape(20484)
Y = X_SD - X_LD
source_space = mne.grade_to_tris(5)
adjacency = mne.spatial_tris_connectivity(source_space)
# adjacency = None
tstep = my_stc_coh_SD[i][w][k][f].tstep
T_obs, clusters, cluster_p_values, H0 = clu = \
mne.stats.permutation_cluster_1samp_test(Y, connectivity=adjacency,
n_jobs=-1,
threshold=t_threshold,
n_permutations=5000,
step_down_p=0.05,
t_power=1, tail=0)
if len(np.where(cluster_p_values < p)[0]) != 0:
print('significant!')
stc_fsaverage.save(stc_fname)
betas.append(stc_fsaverage._data)
all_betas.append(betas)
# convert to np
all_betas = np.array(all_betas, float)
# dimension order should be obs x time x source
all_betas = np.transpose(all_betas, [0, 1, 3, 2])
for ii, regressor in enumerate(var_names):
save_dir = "%s/_RESULTS/%s" % (root_dir, regressor)
# apply the permutation test
connectivity = spatial_tris_connectivity(grade_to_tris(spacing))
t_obs, clusters, cluster_pv, H0 = p1samp(all_betas[:, ii, ...],
threshold=4,
n_permutations=n_perm,
connectivity=connectivity,
buffer_size=None,
verbose=True)
# save result of the cluster test
np.save('%s/_stats_%s.npy' % (save_dir, regressor),
(t_obs, clusters, cluster_pv, H0))
# remove clusters that are too short or too small or not significant
# cluster_bool = threshold_clusters(clusters, cluster_pv, min_times,
# min_sources, min_p)
# save the t-value map as an stc. note: can extract np data with stc._data
tmin, tmax = time_interval
times = mean_stc1.times
mask = (times >= tmin) & (times <= tmax)
X1 = np.mean(X1[:, :, mask], axis=2)[:, :, None]
X2 = np.mean(X2[:, :, mask], axis=2)[:, :, None]
template_stc = copy.deepcopy(template_stc)
template_stc.crop(tmin, tmin + template_stc.tstep)
assert X1.shape == X2.shape
n_samples, n_vertices, n_times = X1.shape
X1 = np.ascontiguousarray(np.swapaxes(X1, 1, 2).reshape(n_samples, -1))
X2 = np.ascontiguousarray(np.swapaxes(X2, 1, 2).reshape(n_samples, -1))
#connectivity = mne.spatio_temporal_src_connectivity(src, n_times)
connectivity = spatial_tris_connectivity(grade_to_tris(5))
for t in thresholds:
from time import time
t0 = time()
T_obs, clusters, cluster_pv, H0 = mem.cache(
permutation_cluster_1samp_test,
ignore=['n_jobs'])(X1 - X2,
threshold=t,
n_permutations=n_permutations,
tail=0,
stat_fun=stat_fun,
connectivity=connectivity,
n_jobs=n_jobs,
seed=0)
print "Time elapsed : %s (s)" % (time() - t0)
def grouplevel_spatial_stats(ListSubj,condition,method,mod,twin,clust_p):
wdir = "/neurospin/meg/meg_tmp/MTT_MEG_Baptiste/MEG"
# load a specific STC morphed on fsaverage to get shape info
stc0_path = (wdir + '/' + ListSubj[0] + '/mne_python/STCS/IcaCorr_' + mod +
'_' + ListSubj[0] + '_' + condition[0] + '_pick_oriNone_' +
method + '_ico-5-fwd-fsaverage.fif-rh.stc')
stc0 = mne.read_source_estimate(stc0_path)
stc0.crop(-0.2,2.5)
ncond = len(condition)
nsub = len(ListSubj)
ntimes = len(stc0.times)
nvertices = stc0.data.shape[0]
# average individual STCs morphed on fsaverage for each cond
AVG_STC_cond = np.empty([nvertices, ntimes, nsub, ncond])
for s,subj in enumerate(ListSubj):
for c,cond in enumerate(condition):
stc_path = (wdir + '/' + subj + '/mne_python/STCS/IcaCorr_' + mod +
'_' + subj + '_' + cond + '_pick_oriNone_' +
method + '_ico-5-fwd-fsaverage.fif-rh.stc')
stc = mne.read_source_estimate(stc_path)
stc.crop(-0.2,2.5)
AVG_STC_cond[:,:,s,c] = stc.data
# optional: restrict computation to temporal window of interest
lower_bound = np.where(stc0.times >= twin[0])[0][0]
upper_bound = np.where(stc0.times >= twin[1])[0][0]
con = mne.spatial_tris_connectivity(grade_to_tris(5))
# array of shapes (obs, time, vertices)
X = []
for c,cond in enumerate(condition):
X.append(np.mean(np.transpose(AVG_STC_cond[:,lower_bound: upper_bound,:,c], [2, 1, 0]),1))
effects = 'A'
factor_levels = [3]
# get f-values only.
def mystat_fun(*args):
return f_mway_rm(np.swapaxes(args, 1, 0), factor_levels=factor_levels,
effects=effects, return_pvals=False)[0]
p_threshold = clust_p
f_threshold = f_threshold_mway_rm(nsub,factor_levels = factor_levels, effects = effects,pvalue= p_threshold)
F_obs, clu, clu_p_val, H0 = mne.stats.permutation_cluster_test(X, threshold=f_threshold,stat_fun=mystat_fun,
connectivity=con, n_jobs=1,
verbose=True, seed = 666)
wdir = "/neurospin/meg/meg_tmp/MTT_MEG_Baptiste/MEG/"
save_path = (wdir+'GROUP/mne_python/Plot_STATS/' + "_vs_".join(condition))
if not os.path.exists(save_path):
os.makedirs(save_path)
# save cluster stats
spatial_clust_F = np.array((F_obs, clu, clu_p_val, H0))
np.save((save_path+'/' + mod + '_' +'cluster_stats_f_'+ "_vs_".join(condition)),
spatial_clust_F)
# save F-Map
tmp = F_obs
tmp = tmp[:,np.newaxis]
fsave_vertices = [np.arange(10242), np.arange(10242)]
stc_Ftest = mne.SourceEstimate(tmp,fsave_vertices,0,stc.tstep)
stc_Ftest.save((save_path + '/fmap' + mod + '_' + "_vs_".join(condition)))
return F_obs, clu, clu_p_val, H0,stc0