Note

Go to the end to download the full example code or to run this example in your browser via Binder.

Comparing connectomes on different reference atlases

This examples shows how to turn a parcellation into connectome for visualization. This requires choosing centers for each parcel or network, via find_parcellation_cut_coords for parcellation based on labels and find_probabilistic_atlas_cut_coords for parcellation based on probabilistic values.

In the intermediary steps, we make use of MultiNiftiLabelsMasker and MultiNiftiMapsMasker to extract time series from nifti objects from multiple subjects using different parcellation atlases.

The time series of all subjects of the brain development dataset are concatenated and given directly to ConnectivityMeasure for computing parcel-wise correlation matrices for each atlas across all subjects.

Mean correlation matrix is displayed on glass brain on extracted coordinates.

Warning

If you are using Nilearn with a version older than 0.9.0, then you should either upgrade your version or import maskers from the input_data module instead of the maskers module.

That is, you should manually replace in the following example all occurrences of:

from nilearn.maskers import NiftiMasker

with:

from nilearn.input_data import NiftiMasker

Load atlases

from nilearn.datasets import fetch_atlas_yeo_2011, fetch_development_fmri

yeo = fetch_atlas_yeo_2011(n_networks=17)
print(
    "Yeo atlas nifti image (3D) with 17 parcels and liberal mask "
    f" is located at: {yeo['maps']}"
)

[fetch_atlas_yeo_2011] Dataset found in /home/runner/nilearn_data/yeo_2011
Yeo atlas nifti image (3D) with 17 parcels and liberal mask  is located at: /home/runner/nilearn_data/yeo_2011/Yeo_JNeurophysiol11_MNI152/Yeo2011_17Networks_MNI152_FreeSurferConformed1mm_LiberalMask.nii.gz

Load functional data

data = fetch_development_fmri(n_subjects=10)

print(
    "Functional nifti images (4D, e.g., one subject) "
    f"are located at : {data.func[0]!r}"
)
print(
    "Counfound csv files (of same subject) are located "
    f"at : {data['confounds'][0]!r}"
)

[fetch_development_fmri] Dataset found in
/home/runner/nilearn_data/development_fmri
[fetch_development_fmri] Dataset found in
/home/runner/nilearn_data/development_fmri/development_fmri
[fetch_development_fmri] Dataset found in
/home/runner/nilearn_data/development_fmri/development_fmri
Functional nifti images (4D, e.g., one subject) are located at : '/home/runner/nilearn_data/development_fmri/development_fmri/sub-pixar123_task-pixar_space-MNI152NLin2009cAsym_desc-preproc_bold.nii.gz'
Counfound csv files (of same subject) are located at : '/home/runner/nilearn_data/development_fmri/development_fmri/sub-pixar123_task-pixar_desc-reducedConfounds_regressors.tsv'

Extract coordinates on Yeo atlas - parcellations

from nilearn.connectome import ConnectivityMeasure
from nilearn.maskers import MultiNiftiLabelsMasker

# ConnectivityMeasure from Nilearn uses simple 'correlation' to compute
# connectivity matrices for all subjects in a list
connectome_measure = ConnectivityMeasure(
    kind="correlation", standardize="zscore_sample", verbose=1
)

# create masker using MultiNiftiLabelsMasker to extract functional data within
# atlas parcels from multiple subjects using parallelization to speed up the
# computation
masker = MultiNiftiLabelsMasker(
    labels_img=yeo["maps"],  # Both hemispheres
    standardize="zscore_sample",
    standardize_confounds=True,
    memory="nilearn_cache",
    n_jobs=2,
    verbose=1,
)

# extract time series from all subjects
time_series = masker.fit_transform(data.func, confounds=data.confounds)

# calculate correlation matrices across subjects and display
correlation_matrices = connectome_measure.fit_transform(time_series)

# Mean correlation matrix across 10 subjects can be grabbed like this,
# using connectome measure object
mean_correlation_matrix = connectome_measure.mean_

# useful for plotting connectivity interactions on glass brain
from nilearn.plotting import (
    find_parcellation_cut_coords,
    plot_connectome,
    show,
)

# grab center coordinates for atlas labels
coordinates = find_parcellation_cut_coords(labels_img=yeo["maps"])

# plot connectome with 80% edge strength in the connectivity
left_connectome = plot_connectome(
    mean_correlation_matrix, coordinates, edge_threshold="80%"
)

show()
plot atlas comparison
[MultiNiftiLabelsMasker.fit_transform] Loading regions from
'/home/runner/nilearn_data/yeo_2011/Yeo_JNeurophysiol11_MNI152/Yeo2011_17Network
s_MNI152_FreeSurferConformed1mm_LiberalMask.nii.gz'
[MultiNiftiLabelsMasker.fit_transform] Resampling regions
[MultiNiftiLabelsMasker.fit_transform] Finished fit
[ConnectivityMeasure.wrapped] Finished fit

Note

The approach above will extract time series and compute a single connectivity matrix for both hemispheres. However, the connectome is plotted only for the left hemisphere.

If your aim is to compute and plot hemisphere-wise connectivity, you can follow the example below.

First, create a separate atlas image for each hemisphere:

import nibabel as nb
import numpy as np

from nilearn.image import get_data, new_img_like
from nilearn.image.resampling import coord_transform

# load the atlas image first
label_image = nb.load(yeo["maps"])

# extract the affine matrix of the image
labels_affine = label_image.affine

# generate image coordinates using affine
x, y, z = coord_transform(0, 0, 0, np.linalg.inv(labels_affine))

# generate an separate image for the left hemisphere
# left/right split is done along x-axis
left_hemi = get_data(label_image).copy()
left_hemi[: int(x)] = 0
label_image_left = new_img_like(label_image, left_hemi, labels_affine)

# same for the right hemisphere
right_hemi = get_data(label_image).copy()
right_hemi[int(x) :] = 0
label_image_right = new_img_like(label_image, right_hemi, labels_affine)

Then, create a masker object, compute a connectivity matrix and plot the results for each hemisphere:

for hemi, img in zip(
    ["right", "left"], [label_image_right, label_image_left], strict=False
):
    masker = MultiNiftiLabelsMasker(
        labels_img=img,
        standardize="zscore_sample",
        standardize_confounds=True,
        verbose=1,
    )

    time_series = masker.fit_transform(data.func, confounds=data.confounds)

    correlation_matrices = connectome_measure.fit_transform(time_series)
    mean_correlation_matrix = connectome_measure.mean_

    coordinates = find_parcellation_cut_coords(
        labels_img=img, label_hemisphere=hemi
    )

    plot_connectome(
        mean_correlation_matrix,
        coordinates,
        edge_threshold="80%",
        title=f"Yeo Atlas 17 thick (func) - {hemi}",
    )

show()
  • plot atlas comparison
  • plot atlas comparison
[MultiNiftiLabelsMasker.fit_transform] Loading regions from
<nibabel.nifti1.Nifti1Image object at 0x7f0f046afa00>
[MultiNiftiLabelsMasker.fit_transform] Resampling regions
[MultiNiftiLabelsMasker.fit_transform] Finished fit
[MultiNiftiLabelsMasker.fit_transform] Loading data from
<nibabel.nifti1.Nifti1Image object at 0x7f0f5653a080>
[MultiNiftiLabelsMasker.fit_transform] Extracting region signals
[MultiNiftiLabelsMasker.fit_transform] Cleaning extracted signals
/home/runner/work/nilearn/nilearn/examples/03_connectivity/plot_atlas_comparison.py:157: DeprecationWarning: From release 0.14.0, confounds will be standardized using the sample std instead of the population std.
  time_series = masker.fit_transform(data.func, confounds=data.confounds)
[MultiNiftiLabelsMasker.fit_transform] Loading data from
<nibabel.nifti1.Nifti1Image object at 0x7f0f5653a560>
[MultiNiftiLabelsMasker.fit_transform] Extracting region signals
[MultiNiftiLabelsMasker.fit_transform] Cleaning extracted signals
/home/runner/work/nilearn/nilearn/examples/03_connectivity/plot_atlas_comparison.py:157: DeprecationWarning: From release 0.14.0, confounds will be standardized using the sample std instead of the population std.
  time_series = masker.fit_transform(data.func, confounds=data.confounds)
[MultiNiftiLabelsMasker.fit_transform] Loading data from
<nibabel.nifti1.Nifti1Image object at 0x7f0f5653a6e0>
[MultiNiftiLabelsMasker.fit_transform] Extracting region signals
[MultiNiftiLabelsMasker.fit_transform] Cleaning extracted signals
/home/runner/work/nilearn/nilearn/examples/03_connectivity/plot_atlas_comparison.py:157: DeprecationWarning: From release 0.14.0, confounds will be standardized using the sample std instead of the population std.
  time_series = masker.fit_transform(data.func, confounds=data.confounds)
[MultiNiftiLabelsMasker.fit_transform] Loading data from
<nibabel.nifti1.Nifti1Image object at 0x7f0f56539cf0>
[MultiNiftiLabelsMasker.fit_transform] Extracting region signals
[MultiNiftiLabelsMasker.fit_transform] Cleaning extracted signals
/home/runner/work/nilearn/nilearn/examples/03_connectivity/plot_atlas_comparison.py:157: DeprecationWarning: From release 0.14.0, confounds will be standardized using the sample std instead of the population std.
  time_series = masker.fit_transform(data.func, confounds=data.confounds)
[MultiNiftiLabelsMasker.fit_transform] Loading data from
<nibabel.nifti1.Nifti1Image object at 0x7f0f56538730>
[MultiNiftiLabelsMasker.fit_transform] Extracting region signals
[MultiNiftiLabelsMasker.fit_transform] Cleaning extracted signals
/home/runner/work/nilearn/nilearn/examples/03_connectivity/plot_atlas_comparison.py:157: DeprecationWarning: From release 0.14.0, confounds will be standardized using the sample std instead of the population std.
  time_series = masker.fit_transform(data.func, confounds=data.confounds)
[MultiNiftiLabelsMasker.fit_transform] Loading data from
<nibabel.nifti1.Nifti1Image object at 0x7f0f5653a5c0>
[MultiNiftiLabelsMasker.fit_transform] Extracting region signals
[MultiNiftiLabelsMasker.fit_transform] Cleaning extracted signals
/home/runner/work/nilearn/nilearn/examples/03_connectivity/plot_atlas_comparison.py:157: DeprecationWarning: From release 0.14.0, confounds will be standardized using the sample std instead of the population std.
  time_series = masker.fit_transform(data.func, confounds=data.confounds)
[MultiNiftiLabelsMasker.fit_transform] Loading data from
<nibabel.nifti1.Nifti1Image object at 0x7f0f5653a230>
[MultiNiftiLabelsMasker.fit_transform] Extracting region signals
[MultiNiftiLabelsMasker.fit_transform] Cleaning extracted signals
/home/runner/work/nilearn/nilearn/examples/03_connectivity/plot_atlas_comparison.py:157: DeprecationWarning: From release 0.14.0, confounds will be standardized using the sample std instead of the population std.
  time_series = masker.fit_transform(data.func, confounds=data.confounds)
[MultiNiftiLabelsMasker.fit_transform] Loading data from
<nibabel.nifti1.Nifti1Image object at 0x7f0f56539570>
[MultiNiftiLabelsMasker.fit_transform] Extracting region signals
[MultiNiftiLabelsMasker.fit_transform] Cleaning extracted signals
/home/runner/work/nilearn/nilearn/examples/03_connectivity/plot_atlas_comparison.py:157: DeprecationWarning: From release 0.14.0, confounds will be standardized using the sample std instead of the population std.
  time_series = masker.fit_transform(data.func, confounds=data.confounds)
[MultiNiftiLabelsMasker.fit_transform] Loading data from
<nibabel.nifti1.Nifti1Image object at 0x7f0f5653b6d0>
[MultiNiftiLabelsMasker.fit_transform] Extracting region signals
[MultiNiftiLabelsMasker.fit_transform] Cleaning extracted signals
/home/runner/work/nilearn/nilearn/examples/03_connectivity/plot_atlas_comparison.py:157: DeprecationWarning: From release 0.14.0, confounds will be standardized using the sample std instead of the population std.
  time_series = masker.fit_transform(data.func, confounds=data.confounds)
[MultiNiftiLabelsMasker.fit_transform] Loading data from
<nibabel.nifti1.Nifti1Image object at 0x7f0f5653a3b0>
[MultiNiftiLabelsMasker.fit_transform] Extracting region signals
[MultiNiftiLabelsMasker.fit_transform] Cleaning extracted signals
/home/runner/work/nilearn/nilearn/examples/03_connectivity/plot_atlas_comparison.py:157: DeprecationWarning: From release 0.14.0, confounds will be standardized using the sample std instead of the population std.
  time_series = masker.fit_transform(data.func, confounds=data.confounds)
[ConnectivityMeasure.wrapped] Finished fit
[MultiNiftiLabelsMasker.fit_transform] Loading regions from
<nibabel.nifti1.Nifti1Image object at 0x7f0f046acca0>
[MultiNiftiLabelsMasker.fit_transform] Resampling regions
[MultiNiftiLabelsMasker.fit_transform] Finished fit
[MultiNiftiLabelsMasker.fit_transform] Loading data from
<nibabel.nifti1.Nifti1Image object at 0x7f0f47723130>
[MultiNiftiLabelsMasker.fit_transform] Extracting region signals
[MultiNiftiLabelsMasker.fit_transform] Cleaning extracted signals
/home/runner/work/nilearn/nilearn/examples/03_connectivity/plot_atlas_comparison.py:157: DeprecationWarning: From release 0.14.0, confounds will be standardized using the sample std instead of the population std.
  time_series = masker.fit_transform(data.func, confounds=data.confounds)
[MultiNiftiLabelsMasker.fit_transform] Loading data from
<nibabel.nifti1.Nifti1Image object at 0x7f0f2f67f0d0>
[MultiNiftiLabelsMasker.fit_transform] Extracting region signals
[MultiNiftiLabelsMasker.fit_transform] Cleaning extracted signals
/home/runner/work/nilearn/nilearn/examples/03_connectivity/plot_atlas_comparison.py:157: DeprecationWarning: From release 0.14.0, confounds will be standardized using the sample std instead of the population std.
  time_series = masker.fit_transform(data.func, confounds=data.confounds)
[MultiNiftiLabelsMasker.fit_transform] Loading data from
<nibabel.nifti1.Nifti1Image object at 0x7f0f47722e30>
[MultiNiftiLabelsMasker.fit_transform] Extracting region signals
[MultiNiftiLabelsMasker.fit_transform] Cleaning extracted signals
/home/runner/work/nilearn/nilearn/examples/03_connectivity/plot_atlas_comparison.py:157: DeprecationWarning: From release 0.14.0, confounds will be standardized using the sample std instead of the population std.
  time_series = masker.fit_transform(data.func, confounds=data.confounds)
[MultiNiftiLabelsMasker.fit_transform] Loading data from
<nibabel.nifti1.Nifti1Image object at 0x7f0f47722b90>
[MultiNiftiLabelsMasker.fit_transform] Extracting region signals
[MultiNiftiLabelsMasker.fit_transform] Cleaning extracted signals
/home/runner/work/nilearn/nilearn/examples/03_connectivity/plot_atlas_comparison.py:157: DeprecationWarning: From release 0.14.0, confounds will be standardized using the sample std instead of the population std.
  time_series = masker.fit_transform(data.func, confounds=data.confounds)
[MultiNiftiLabelsMasker.fit_transform] Loading data from
<nibabel.nifti1.Nifti1Image object at 0x7f0f47722ad0>
[MultiNiftiLabelsMasker.fit_transform] Extracting region signals
[MultiNiftiLabelsMasker.fit_transform] Cleaning extracted signals
/home/runner/work/nilearn/nilearn/examples/03_connectivity/plot_atlas_comparison.py:157: DeprecationWarning: From release 0.14.0, confounds will be standardized using the sample std instead of the population std.
  time_series = masker.fit_transform(data.func, confounds=data.confounds)
[MultiNiftiLabelsMasker.fit_transform] Loading data from
<nibabel.nifti1.Nifti1Image object at 0x7f0f47722dd0>
[MultiNiftiLabelsMasker.fit_transform] Extracting region signals
[MultiNiftiLabelsMasker.fit_transform] Cleaning extracted signals
/home/runner/work/nilearn/nilearn/examples/03_connectivity/plot_atlas_comparison.py:157: DeprecationWarning: From release 0.14.0, confounds will be standardized using the sample std instead of the population std.
  time_series = masker.fit_transform(data.func, confounds=data.confounds)
[MultiNiftiLabelsMasker.fit_transform] Loading data from
<nibabel.nifti1.Nifti1Image object at 0x7f0f47722950>
[MultiNiftiLabelsMasker.fit_transform] Extracting region signals
[MultiNiftiLabelsMasker.fit_transform] Cleaning extracted signals
/home/runner/work/nilearn/nilearn/examples/03_connectivity/plot_atlas_comparison.py:157: DeprecationWarning: From release 0.14.0, confounds will be standardized using the sample std instead of the population std.
  time_series = masker.fit_transform(data.func, confounds=data.confounds)
[MultiNiftiLabelsMasker.fit_transform] Loading data from
<nibabel.nifti1.Nifti1Image object at 0x7f0f2f67e5f0>
[MultiNiftiLabelsMasker.fit_transform] Extracting region signals
[MultiNiftiLabelsMasker.fit_transform] Cleaning extracted signals
/home/runner/work/nilearn/nilearn/examples/03_connectivity/plot_atlas_comparison.py:157: DeprecationWarning: From release 0.14.0, confounds will be standardized using the sample std instead of the population std.
  time_series = masker.fit_transform(data.func, confounds=data.confounds)
[MultiNiftiLabelsMasker.fit_transform] Loading data from
<nibabel.nifti1.Nifti1Image object at 0x7f0f47722e00>
[MultiNiftiLabelsMasker.fit_transform] Extracting region signals
[MultiNiftiLabelsMasker.fit_transform] Cleaning extracted signals
/home/runner/work/nilearn/nilearn/examples/03_connectivity/plot_atlas_comparison.py:157: DeprecationWarning: From release 0.14.0, confounds will be standardized using the sample std instead of the population std.
  time_series = masker.fit_transform(data.func, confounds=data.confounds)
[MultiNiftiLabelsMasker.fit_transform] Loading data from
<nibabel.nifti1.Nifti1Image object at 0x7f0f2f67e5f0>
[MultiNiftiLabelsMasker.fit_transform] Extracting region signals
[MultiNiftiLabelsMasker.fit_transform] Cleaning extracted signals
/home/runner/work/nilearn/nilearn/examples/03_connectivity/plot_atlas_comparison.py:157: DeprecationWarning: From release 0.14.0, confounds will be standardized using the sample std instead of the population std.
  time_series = masker.fit_transform(data.func, confounds=data.confounds)
[ConnectivityMeasure.wrapped] Finished fit

Plot a directed connectome - asymmetric connectivity measure

In this section, we use the lag-1 correlation as the connectivity measure, which leads to an asymmetric connectivity matrix. The plot_connectome function accepts both symmetric and asymmetric matrices, but plots the latter as a directed graph.

# Define a custom function to compute lag correlation on the time series
def lag_correlation(time_series, lag):
    n_subjects = len(time_series)
    _, n_features = time_series[0].shape
    lag_cor = np.zeros((n_subjects, n_features, n_features))
    for subject, serie in enumerate(time_series):
        for i in range(n_features):
            for j in range(n_features):
                if lag == 0:
                    lag_cor[subject, i, j] = np.corrcoef(
                        serie[:, i], serie[:, j]
                    )[0, 1]
                else:
                    lag_cor[subject, i, j] = np.corrcoef(
                        serie[lag:, i], serie[:-lag, j]
                    )[0, 1]
    return np.mean(lag_cor, axis=0)


# Compute lag-0 and lag-1 correlations and plot associated connectomes
for lag in [0, 1]:
    lag_correlation_matrix = lag_correlation(time_series, lag)
    plot_connectome(
        lag_correlation_matrix,
        coordinates,
        edge_threshold="90%",
        title=f"Lag-{lag} correlation",
    )
  • plot atlas comparison
  • plot atlas comparison
/home/runner/work/nilearn/nilearn/examples/03_connectivity/plot_atlas_comparison.py:206: UserWarning: 'adjacency_matrix' is not symmetric.
A directed graph will be plotted.
  plot_connectome(

Load probabilistic atlases - extracting coordinates on brain maps

from nilearn.datasets import fetch_atlas_difumo
from nilearn.plotting import find_probabilistic_atlas_cut_coords

dim = 64
difumo = fetch_atlas_difumo(dimension=dim, resolution_mm=2)

[fetch_atlas_difumo] Dataset found in /home/runner/nilearn_data/difumo_atlases

Iterate over fetched atlases to extract coordinates - probabilistic

from nilearn.maskers import MultiNiftiMapsMasker

# Create masker using MultiNiftiMapsMasker to extract functional data within
# atlas parcels from multiple subjects using parallelization to speed up the
# computation.
masker = MultiNiftiMapsMasker(
    maps_img=difumo.maps,
    standardize="zscore_sample",
    standardize_confounds=True,
    memory="nilearn_cache",
    memory_level=1,
    n_jobs=2,
    verbose=1,
)

# extract time series from all subjects
time_series = masker.fit_transform(data.func, confounds=data.confounds)

# calculate correlation matrices across subjects and display
correlation_matrices = connectome_measure.fit_transform(time_series)

# Mean correlation matrix across 10 subjects can be grabbed like this,
# using connectome measure object
mean_correlation_matrix = connectome_measure.mean_

# grab center coordinates for probabilistic atlas
coordinates = find_probabilistic_atlas_cut_coords(maps_img=difumo.maps)

# plot connectome with 85% edge strength in the connectivity
plot_connectome(
    mean_correlation_matrix,
    coordinates,
    edge_threshold="85%",
    title=f"DiFuMo with {dim} dimensions (probabilistic)",
)
show()
plot atlas comparison
[MultiNiftiMapsMasker.fit_transform] Loading regions from
'/home/runner/nilearn_data/difumo_atlases/64/2mm/maps.nii.gz'
[MultiNiftiMapsMasker.fit_transform] Resampling regions
________________________________________________________________________________
[Memory] Calling nilearn.image.resampling.resample_img...
resample_img(<nibabel.nifti1.Nifti1Image object at 0x7f0f0c4cf0a0>, interpolation='linear', target_shape=(50, 59, 50), target_affine=array([[   4.,    0.,    0.,  -96.],
       [   0.,    4.,    0., -132.],
       [   0.,    0.,    4.,  -78.],
       [   0.,    0.,    0.,    1.]]))
_____________________________________________________resample_img - 0.7s, 0.0min
[MultiNiftiMapsMasker.fit_transform] Finished fit
[ConnectivityMeasure.wrapped] Finished fit

Total running time of the script: (2 minutes 0.110 seconds)

Estimated memory usage: 3587 MB

Gallery generated by Sphinx-Gallery