Real fMRI data example

Example to recover the different neural temporal activities, the associated functional networks maps and the HRFs per ROIs in the fMRI data, on the ADHD dataset resting-state.

# Authors: Hamza Cherkaoui <hamza.cherkaoui@inria.fr>
# License: BSD (3-clause)

import os
from nilearn import datasets
import numpy as np

from hemolearn import BDA, plotting

Create plotting directory

plot_dir = 'plots'
if not os.path.exists(plot_dir):
    os.makedirs(plot_dir)

Fetch fMRI subjects

seed, TR, n_subjects = 0, 2.0, 4
adhd_dataset = datasets.fetch_adhd(n_subjects=n_subjects)
func_fnames = adhd_dataset.func[:n_subjects]
confound_fnames = adhd_dataset.confounds[:n_subjects]

Out:

/home/hcherkaoui/.local/lib/python3.6/site-packages/nilearn/datasets/func.py:241: VisibleDeprecationWarning: Reading unicode strings without specifying the encoding argument is deprecated. Set the encoding, use None for the system default.
  dtype=None)

Distangle the neurovascular coupling from the neural activation

bda = BDA(n_atoms=10, t_r=TR, n_times_atom=30, lbda=0.1, max_iter=50,
          standardize=True, shared_spatial_maps=True, random_state=seed,
          verbose=1)
a_hat_img = bda.fit_transform(func_fnames, confound_fnames=confound_fnames)

Out:

[BDA] Clean subject '/home/hcherkaoui/nilearn_data/adhd/data/0010042/0010042_rest_tshift_RPI_voreg_mni.nii.gz'
[BDA] Clean subject '/home/hcherkaoui/nilearn_data/adhd/data/0010064/0010064_rest_tshift_RPI_voreg_mni.nii.gz'
[BDA] Clean subject '/home/hcherkaoui/nilearn_data/adhd/data/0010128/0010128_rest_tshift_RPI_voreg_mni.nii.gz'
[BDA] Clean subject '/home/hcherkaoui/nilearn_data/adhd/data/0021019/0021019_rest_tshift_RPI_voreg_mni.nii.gz'
[BDA] Running 1 fit(s) on 4 subject(s) in series

Plot the spatial maps

filename = os.path.join(plot_dir, f'spatial_maps.png')
plotting.plot_spatial_maps(bda.u_hat_img, filename=filename,
                           perc_voxels_to_retain='1%', verbose=True)

Out:

Saving plot under 'plots/spatial_maps.png'

Plot the temporal activations

for n in range(n_subjects):
    filename = os.path.join(plot_dir, f'activations_{n}.png')
    plotting.plot_temporal_activations(bda.z_hat[n], TR, filename=filename,
                                       verbose=True)

• 
• 
• 
• 

Out:

Saving plot under 'plots/activations_0.png'
Saving plot under 'plots/activations_1.png'
Saving plot under 'plots/activations_2.png'
Saving plot under 'plots/activations_3.png'

Plot vascular maps

vmax = np.max([np.max(bda.a_hat[n]) for n in range(n_subjects)])
for n in range(n_subjects):
    filename = os.path.join(plot_dir, f'vascular_maps_{n}.png')
    plotting.plot_vascular_map(a_hat_img[n], display_mode='z',
                               cut_coords=np.linspace(-30, 60, 5),
                               filename=filename, vmax=vmax, verbose=True)

• 
• 
• 
• 

Out:

/home/hcherkaoui/.local/lib/python3.6/site-packages/nilearn/plotting/img_plotting.py:341: FutureWarning: Default resolution of the MNI template will change from 2mm to 1mm in version 0.10.0
  anat_img = load_mni152_template()
Saving plot under 'plots/vascular_maps_0.png'
Saving plot under 'plots/vascular_maps_1.png'
Saving plot under 'plots/vascular_maps_2.png'
Saving plot under 'plots/vascular_maps_3.png'