def get_X_y(coordinate_file, feature_file, filter=True):
if filter:
coordinate_dict = ex.filter_studies_active_voxels(coordinate_file)
else:
with open(coordinate_file) as f:
coordinate_dict = json.load(f)
target_dict = ex.filter_studies_terms(feature_file, set_unique_label=True)
coordinate_dict, target_dict = ex.get_intersecting_dicts(coordinate_dict,
target_dict)
X, y = pp.get_features_targets(coordinate_dict, target_dict)
return X, y
def main():
with open('data/docdict.txt') as f:
coordinates = json.load(f)
study_dict = ex.filter_studies_active_voxels(coordinates, 'data/MNI152_T1_2mm_brain.nii.gz', radius=6, threshold=500)
# ensure that study dict has int as keys
for key in list(study_dict):
study_dict[int(key)] = study_dict[key]
if not isinstance(key, int):
del(study_dict[key])
with open('data/terms.json', 'rb') as f:
terms = json.load(f)
feature_dict = ex.filter_studies_terms('data/features.txt', terms=terms, set_unique_label=True)
for i in range(len(terms)):
for j in range(i+1, len(terms)):
sub_dict = {}
for key in feature_dict:
if feature_dict[key] in [terms[i], terms[j]]:
sub_dict[key] = feature_dict[key]
study, feat = ex.get_intersecting_dicts(study_dict, sub_dict)
x, y = pp.get_features_targets(study, feat, mask='data/MNI152_T1_2mm_brain.nii.gz')
cf = classify(x, y)
save_name = terms[i] + '_vs_' + terms[j] + '.npy'
with open(save_name, 'wb') as f:
np.save(f, cf)
