def download_subset():
data_dir = get_data_dirs()[0]
error_dir = join(data_dir, 'failures')
if not os.path.exists(error_dir):
os.makedirs(error_dir)
n_jobs = 4
subjects = fetch_subject_list()[:4]
Parallel(n_jobs=n_jobs,
verbose=10)(delayed(download_single)(subject, verbose=2)
for subject in subjects)
def contrasts():
data_dir = get_data_dirs()[0]
error_dir = join(data_dir, 'failures')
if not os.path.exists(error_dir):
os.makedirs(error_dir)
n_jobs = 36
subjects = fetch_subject_list()
tasks = TASK_LIST
Parallel(n_jobs=n_jobs, verbose=10)(
delayed(make_contrasts)(subject, task, overwrite=True, verbose=1)
for subject in subjects for task in tasks)
from hcp_builder.dataset import fetch_subject_list
from hcp_builder.utils.fsl import clean_artifacts
subject_list = fetch_subject_list()
total = len(subject_list)
for i, subject in enumerate(subject_list):
print('Clearning subject %s, %i / %i' % (subject, i, total))
clean_artifacts(subject)
import numpy as np
from joblib import Parallel, delayed, Memory
import matplotlib.pyplot as plt
from nilearn.input_data import NiftiMasker
from pybold.bold_signal import bd
from pybold.hrf_model import spm_hrf, MIN_DELTA, MAX_DELTA
from pybold.utils import inf_norm
from hrf_estimation.rank_one_ import glm as pglm
from hcp_builder.dataset import fetch_subject_list
from utils import (create_result_dir, get_hcp_fmri_fname, get_protocol_hcp,
mask_n_max, plot_trial_z_map, hrf_coef_to_hrf, TR_HCP,
N_SCANS_HCP, DUR_RUN_HCP)
# usefull functions
if not fetch_subject_list():
def fetch_subject_list():
"""Fix troubles cause by bug/wrong usage with hcp_builder."""
return [100206, 996782]
pglm_cached = Memory('./.cachedir').cache(pglm, ignore=['n_jobs', 'verbose'])
def _bd(voxels, lbda, hrf_dur, n_jobs, verbose=10):
"""Blind deconvolution"""
res = Parallel(n_jobs=n_jobs,
verbose=verbose)(delayed(bd)(voxel,
t_r=TR_HCP,
lbda=lbda,