def perform_analysis():
results = []
for subj in subj_list:
labels = [
mtk.get_behavior_data(behav_dir, subj, run, label) for run in runs
]
imgs = [
mtk.masking_fmri_image(img_data[(subj, run)], mask) for run in runs
]
data_xs = np.concatenate(imgs)
data_ys = list(labels[0]['task_type']) + list(labels[1]['task_type'])
group = [1 for _ in labels[0]['task_type']
] + [2 for _ in labels[1]['task_type']]
cv = GroupKFold(n_splits=2)
cv_scores = cross_val_score(estimator,
data_xs,
data_ys,
cv=cv,
groups=group)
results.append(list(cv_scores))
return results
def initial_images(standardize_trial=False):
result = {}
for run in runs:
for subj in subj_list:
labels = mtk.get_behavior_data(behav_dir, subj, run, 'color')
# load and resampling image
img = mtk.load_5d_fmri_image(data_dir + 'tvalsLSA.%s.r0%d.nii.gz' %
(subj, run))
img = nilearn.image.index_img(img, labels['order'] - 1)
if standardize_trial:
temp = mtk.masking_fmri_image(img, mask_img)
temp = temp - np.tile(np.mean(temp, axis=1),
(temp.shape[-1], 1)).T
img = nilearn.masking.unmask(temp, mask_img)
img = nilearn.image.resample_img(img,
roi_masks[0].affine,
roi_masks[0].shape,
interpolation='nearest')
result[subj, run] = img
return result
def perform_analysis():
results = []
for subj in subj_list:
labels = mtk.get_behavior_data(behav_dir, subj, run, label)
imgs = mtk.masking_fmri_image(img_data[subj], mask)
data_xs = imgs
data_ys = list(labels['task_type'])
cv = mtk.BalancedShuffleSplit(n_splits=2)
cv_scores = cross_val_score(estimator, data_xs, data_ys, cv=cv)
results.append(list(cv_scores))
return results
def _perform_analysis(subj, estimator, run, label, niter, standardize=True):
if estimator == 'gnb':
estimator = GaussianNB()
elif estimator == 'svc':
estimator = Pipeline([('scale', StandardScaler()),
('svc', LinearSVC())])
# load behavioral data
labels = get_behavior_data(behav_dir, subj, run, label)
# load fmri data
img = load_fmri_image(data_dir, subj, run, labels, standardize=standardize)
y = labels['task_type']
return run_roi_based_mvpa(estimator,
img,
y,
roi_masks,
'balanced',
n_iter=niter)
def _perform_analysis(subj, estimator, run, label):
if estimator == 'gnb':
estimator = GaussianNB()
elif estimator == 'svc':
estimator = Pipeline([('scale', StandardScaler()),
('svc', LinearSVC())])
# load behavioral data
labels = get_behavior_data(behav_dir,
subj,
run,
label,
contain_groups=(1, ))
# load fmri data
img = load_fmri_image(data_dir, subj, run, labels)
y = labels['task_type']
return labels['order'], run_decoding_time_series(estimator, img, y,
roi_masks)
'color': [3, 4],
}
runs = run_number_dict[label]
# load mask file
mask_img = get_full_mask(data_dir)
for subj in subj_list:
print('starting run %s, %s label' % (subj, label))
for run in runs:
# load behavioral data
labels = get_behavior_data(behav_dir,
subj,
run,
label,
stratified_group=True)
# load fmri file
img = load_fmri_image(data_dir, subj, run, labels)
X = img
y = list(labels['task_type'])
group = list(labels['group'])
searchlight_img = run_searchlight(mask_img,
X,
y,
group,
estimator,
run_number_dict = {
'move': [3, 5],
'plan': [3, 4],
'color': [3, 4],
}
runs = run_number_dict[label]
# load mask file
mask_img = mtk.get_full_mask(mask_dir)
for subj in subj_list:
print('starting run %s, %s label' % (subj, label))
labels = [
mtk.get_behavior_data(behav_dir, subj, run, label)
for run in runs
]
imgs = [
nilearn.image.index_img(
mtk.load_5d_fmri_image(data_dir + 'tvalsLSA.%s.r0%d.nii.gz' % (subj, run)),
label['order'] - 1)
for label, run in zip(labels, runs)
]
X = nilearn.image.concat_imgs(imgs)
y = list(labels[0]['task_type']) + list(labels[1]['task_type'])
group = [1 for _ in labels[0]['task_type']] + [2 for _ in labels[1]['task_type']]
searchlight_img = mtk.run_searchlight(mask_img, X, y, group, group_k=2,