Hello MNE community,
I am working with source localization in surface EEG. I used LCMV and eLORETA and want to compare two specific time points of the Source Estimates between two conditions. I explored different alternatives and decided to use spatiotemporal clusters ( spatio_temporal_cluster_1samp_test) to assess the difference between the conditions. I am only interested in spatial differences, so I discarded the time axis and shaped my data as (subjects × time × sources loc) = 23 × 1 × 81924.
So, I was wondering if there are any restrictions on using the spatio_temporal_cluster functions. I was able to run my code, but I see the brain empty (white), so I am not sure whether there are truly no clusters, there is a problem in the way I plotted, or if I am making a conceptual error and should use another approach/function. Here is my code.
def statistics_source_space(list_files_per_condition, save_dir_averages, method, time_points, n_permutations, lowpass, overwrite):
input_data = {‘condition1’: [], ‘condition2’:[]}
for file in list_files_per_condition:
stc = io.read_source_estimates(file, ‘’, lowpass, method, morphed=True)
vertices = stc.vertices # store vertices once ideally
if ‘condition1’ in file:
idx = stc.time_as_index(time_points[‘condition1’])
input_data[‘condition1’].append(stc.data[:, idx])
elif ‘condition2’ in file:
idx =stc.time_as_index(time_points[‘condition2’])
input_data[‘condition2’].append(stc.data[:, idx])
read_fsaverage_dir = r’C:\Sourcelocalization\data\grand_averages’
forward = io.read_forward(‘fsaverage.fif’, read_fsaverage_dir)
adjacency = spatial_src_adjacency(forward[‘src’])
n_subjects = len(input_data[‘condition1’])
#Convert to arrays with shape (subjects, 1,vertices) for cluster test
cond1_data = np.stack(input_data[‘condition1’], axis=0)
cond2_data = np.stack(input_data[‘condition2’], axis=0)
X = cond1_data - cond2_data
X = np.transpose(X, [0, 2, 1]) # subject x time x source (size = 23 x 1 x 81924)
#Set up cluster analysis
p_threshold = 0.05
t_threshold = -stats.distributions.t.ppf(p_threshold / 2, n_subjects - 1)
seed = 10
print(‘Clustering.’)
#Cluster permutation 1000 permutations
T_obs, clusters, cluster_p_values, H0 = clu = mne.stats.spatio_temporal_cluster_1samp_test(X, adjacency=adjacency,n_permutations=n_permutations,threshold=t_threshold, seed=seed, n_jobs=12)
stc_all_cluster_vis = summarize_clusters_stc(clu, vertices=vertices,subject=‘fsaverage’)
colormap = ‘hot’
brain = stc_all_cluster_vis.plot(hemi=‘both’, views=‘lat’, colormap=colormap,size=(600, 600), background=‘w’, colorbar=False,smoothing_steps=3)
This is my output
This is the resulting plot
Many thanks,
Christian