def debiasing(data_file, mask, mtx, idx_u, idx_v, tr, out_dir, history_str):
"""
Perform debiasing based on denoised edge-time matrix.
"""
print("Performing debiasing based on denoised edge-time matrix...")
masker = NiftiLabelsMasker(
labels_img=mask,
standardize=False,
memory="nilearn_cache",
strategy="mean",
)
# Read data
data = masker.fit_transform(data_file)
# Generate mask of significant edge-time connections
ets_mask = np.zeros(data.shape)
idxs = np.where(mtx != 0)
time_idxs = idxs[0]
edge_idxs = idxs[1]
print("Generating mask of significant edge-time connections...")
for idx, time_idx in enumerate(time_idxs):
ets_mask[time_idx, idx_u[edge_idxs[idx]]] = 1
ets_mask[time_idx, idx_v[edge_idxs[idx]]] = 1
# Create HRF matrix
hrf = HRFMatrix(
TR=tr,
TE=[0],
nscans=data.shape[0],
r2only=True,
has_integrator=False,
is_afni=True,
)
hrf.generate_hrf()
# Perform debiasing
deb_output = debiasing_spike(hrf, data, ets_mask)
beta = deb_output["beta"]
fitt = deb_output["betafitts"]
# Transform results back to 4D
beta_4D = masker.inverse_transform(beta)
beta_file = join(out_dir, f"{basename(data_file[:-7])}_beta_ETS.nii.gz")
beta_4D.to_filename(beta_file)
atlas_mod.inverse_transform(beta_file, data_file)
subprocess.run(f"3dNotes {join(out_dir, beta_file)} -h {history_str}",
shell=True)
fitt_4D = masker.inverse_transform(fitt)
fitt_file = join(out_dir, f"{basename(data_file[:-7])}_fitt_ETS.nii.gz")
fitt_4D.to_filename(fitt_file)
subprocess.run(f"3dNotes {join(out_dir, fitt_file)} -h {history_str}",
shell=True)
atlas_mod.inverse_transform(fitt_file, data_file)
print("Debiasing finished and files saved.")
return beta, fitt
config = json.load(open(join(artifact_dir, 'config.json'), 'r'))
model = load(join(artifact_dir, 'estimator.pkl'))
maps = model.coef_
source = config['source']
if source == 'craddock':
components = fetch_craddock_parcellation().parcellate400
data = np.ones_like(check_niimg(components).get_data())
mask = new_img_like(components, data)
label_masker = NiftiLabelsMasker(labels_img=components,
smoothing_fwhm=0,
mask_img=mask).fit()
maps_img = label_masker.inverse_transform(maps)
else:
mask = fetch_mask()
masker = MultiNiftiMasker(mask_img=mask).fit()
if source == 'msdl':
components = fetch_atlas_msdl()['maps']
components = masker.transform(components)
elif source in ['hcp_rs', 'hcp_rs_concat', 'hcp_rs_positive']:
data = fetch_atlas_modl()
if source == 'hcp_rs':
components_imgs = [data.nips2017_components64]
elif source == 'hcp_rs_concat':
components_imgs = [
data.nips2017_components16, data.nips2017_components64,
data.nips2017_components256
cur_pval = (1. + np.sum(perm_Rs[1:, 0] > actual_Rs[i_coef])) / n_permutations
pvals.append(cur_pval)
print(cur_pval)
pvals = np.array(pvals)
print('%i CCs are significant at p<0.05' % np.sum(pvals < 0.05))
print('%i CCs are significant at p<0.01' % np.sum(pvals < 0.01))
print('%i CCs are significant at p<0.001' % np.sum(pvals < 0.001))
################
## plot data: ##
################
AAL_weights = pca_brain.inverse_transform(actual_cca.y_weights_[:, 0])
out_nii_CCA = masker.inverse_transform(AAL_weights[None, :])
out_nii_CCA.to_filename('out_nii_CCA{}.nii'.format(1))
out_beh_weights = pca_beh.inverse_transform(actual_cca.x_weights_[:, 0])
out_beh_weights = out_beh_weights.reshape((36, 1))
# save data to dataframe for later analyses
original_beh = {}
original_brain = {}
original_beh["beh_original"] = np.squeeze(out_beh_weights)
original_brain["brain_original"] = AAL_weights
df_brain_vo = pd.DataFrame(data = original_brain)
df_beh_vo = pd.DataFrame(data = original_beh)
df_brain_vo.to_pickle("df_brain_vo")
df_beh_vo["variables"] = df_beh.columns
df_beh_vo = df_beh_vo.set_index("variables")
# 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)
coef_hand = pipeline.named_steps['svm'].coef_
weight_h = masker.inverse_transform(coef_hand)
plot_stat_map(weight_h, title='Train Imp',display_mode='z',cmap='bwr',threshold=0.1)
from sklearn.preprocessing import normalize
#nh=normalize(coef_imp)
#nf=normalize(coef_imag)
#coef_common=nh.T*nf.T
coef_common=coef_hand.T*coef_foot.T
z = coef_common[:]
print("Plot mean activation (per condition) for the first two ROIs.")
print("Use the new ROIs to extract data maps in both ROIs.")
first_roi_masker = NiftiLabelsMasker(labels_img=first_roi_img,
resampling_target=None,
standardize=False, detrend=False)
first_roi_masker.fit()
second_roi_masker = NiftiLabelsMasker(labels_img=second_roi_img,
resampling_target=None,
standardize=False, detrend=False)
second_roi_masker.fit()
condition_names = list(set(haxby_labels))
n_cond_img = masked_fmri_vectors[haxby_labels == 'house', :].shape[0]
n_conds = len(condition_names)
for i, cond in enumerate(condition_names):
cond_mask = haxby_labels == cond
cond_img = image.index_img(fmri_masked_img, cond_mask)
X1_vectors = mem.cache(first_roi_masker.transform)(cond_img).T
X2_vectors = mem.cache(second_roi_masker.transform)(cond_img).T
X1_image = first_roi_masker.inverse_transform(X1_vectors)
X2_image = second_roi_masker.inverse_transform(X2_vectors)
X1_mean_img = mem.cache(image.mean_img)(X1_image)
X2_mean_img = mem.cache(image.mean_img)(X2_image)
plot_two_maps(plot_stat_map,
(X1_mean_img, 'ROI #1 mean %s image' % cond),
(X2_mean_img, 'ROI #2 mean %s image' % cond))
plt.show()
print("Compare activation profiles for face vs. house, for the first two ROIs.")
# Keep the proba of being hetero according to this fit, for the level 1 LogReg stalking
probs = np.concatenate((probs, proba[:, 1] ), axis=None) # kept all test set folds
# Keep this for the LogReg stalking
output_level_0.append(probs) # 86 probas for 100 ROIs
# for checking model fit quality
accs_level_0.append(np.mean(sample_accs))
accs_std_level_0.append(np.std(sample_accs))
# arraying the results
accs_level_0 = np.array(accs_level_0)
accs_std_level_0 = np.array(accs_std_level_0)
# niftiing the accuracies results
level0_accs4D = np.array([accs_level_0]*121) # set as fake 4D for nii transform
accs_level0_nii4D = masker.inverse_transform(level0_accs4D)
accs_level0_nii3D = image.index_img(accs_level0_nii4D, 0)
accs_level0_nii3D.to_filename("accs_level0_3D.nii") # transform as nii and save
# niftiing the accuracies stand deviation results
level0_accs_std4D = np.array([accs_std_level_0]*121) # set as fake 4D for nii transform
accs_std_level0_nii4D = masker.inverse_transform(level0_accs_std4D)
accs_std_level0_nii3D = image.index_img(accs_std_level0_nii4D, 0)
accs_std_level0_nii3D.to_filename("accs_std_level0_3D.nii") # transform as nii and save
# save results as csv to check when rerun the script #reproducibility
coefs_to_save = np.array([accs_level_0, accs_std_level_0]).T
df_accs_per_roi_level0 = pd.DataFrame(data=coefs_to_save, columns=["Accs_level0", "Accs_level0_std"], index=atlas.labels)
df_accs_per_roi_level0.to_excel("accs_per_roi_level0.xlsx")
final_coefficients = np.mean(sample_coefs, axis=0)
final_coefficients_std = np.std(sample_coefs, axis=0)
# format results for the confusion matrix
all_y_pred_ = np.squeeze(np.array(all_y_pred).reshape(1, 4500))
all_y_true_ = np.squeeze(np.array(all_y_true).reshape(1, 4500))
# save results as csv to check when rerun the script #reproducibility
coefs_to_save = np.array([final_coefficients, final_coefficients_std]).T
df_coef_per_roi_mri = pd.DataFrame(data=coefs_to_save,
columns=["Coef", "Coef_std"],
index=atlas.labels)
df_coef_per_roi_mri.to_excel("coef_per_roi_mri.xlsx")
# niftiing the MRI results
final_coefficients_nii3D = masker.inverse_transform(
final_coefficients.reshape(1, 100))
final_coefficients_nii3D.to_filename(
"final_coefficients3D_mri.nii") # transform as nii and save
final_coefficients_std_nii3D = masker.inverse_transform(
final_coefficients_std.reshape(1, 100))
final_coefficients_std_nii3D.to_filename(
"final_coefficients_std3D_mri.nii") # transform as nii and save
################################
#### CONFUSION MATRIX ##########
################################
def rotateTickLabels(ax, rotation, which, rotation_mode='anchor', ha='left'):
''' Plotting function for the x axis labels to be centered
