Seed-based connectivity on the surface

The dataset that is a subset of the enhanced NKI Rockland sample (https://fcon_1000.projects.nitrc.org/indi/enhanced/, Nooner et al.[1])

Resting state fMRI scans (TR=645ms) of 102 subjects were preprocessed (https://github.com/fliem/nki_nilearn) and projected onto the Freesurfer fsaverage5 template (Dale et al.[2] and Fischl et al.[3]). For this example we use the time series of a single subject’s left hemisphere.

The Destrieux parcellation (Destrieux et al.[4]) in fsaverage5 space as distributed with Freesurfer is used to select a seed region in the posterior cingulate cortex.

Functional connectivity of the seed region to all other cortical nodes in the same hemisphere is calculated using Pearson product-moment correlation coefficient.

The nilearn.plotting.plot_surf_stat_map function is used to plot the resulting statistical map on the (inflated) pial surface.

See also for a similar example but using volumetric input data.

See Plotting brain images for more details on plotting tools.

Retrieving the data

# NKI resting state data from nilearn
from nilearn import datasets

nki_dataset = datasets.fetch_surf_nki_enhanced(n_subjects=1)

# The nki dictionary contains file names for the data
# of all downloaded subjects.
print(
    "Resting state data of the first subjects on the "
    f"fsaverag5 surface left hemisphere is at: {nki_dataset['func_left'][0]}"
)

# Destrieux parcellation for left hemisphere in fsaverage5 space
destrieux_atlas = datasets.fetch_atlas_surf_destrieux()
parcellation = destrieux_atlas["map_left"]
labels = destrieux_atlas["labels"]

# Fsaverage5 surface template
fsaverage = datasets.fetch_surf_fsaverage()

# The fsaverage dataset contains file names pointing to
# the file locations
print(
    "Fsaverage5 pial surface of left hemisphere is at: "
    f"{fsaverage['pial_left']}"
)
print(
    "Fsaverage5 flatten pial surface of left hemisphere is at: "
    f"{fsaverage['flat_left']}"
)
print(
    "Fsaverage5 inflated surface of left hemisphere is at: "
    f"{fsaverage['infl_left']}"
)
print(
    "Fsaverage5 sulcal depth map of left hemisphere is at: "
    f"{fsaverage['sulc_left']}"
)
print(
    "Fsaverage5 curvature map of left hemisphere is at: "
    f"{fsaverage['curv_left']}"
)

[get_dataset_dir] Dataset found in /home/runner/work/nilearn/nilearn/nilearn_data/nki_enhanced_surface
Resting state data of the first subjects on the fsaverag5 surface left hemisphere is at: /home/runner/work/nilearn/nilearn/nilearn_data/nki_enhanced_surface/A00028185/A00028185_left_preprocessed_fwhm6.gii
[get_dataset_dir] Dataset found in /home/runner/work/nilearn/nilearn/nilearn_data/destrieux_surface
Fsaverage5 pial surface of left hemisphere is at: /opt/hostedtoolcache/Python/3.12.4/x64/lib/python3.12/site-packages/nilearn/datasets/data/fsaverage5/pial_left.gii.gz
Fsaverage5 flatten pial surface of left hemisphere is at: /opt/hostedtoolcache/Python/3.12.4/x64/lib/python3.12/site-packages/nilearn/datasets/data/fsaverage5/flat_left.gii.gz
Fsaverage5 inflated surface of left hemisphere is at: /opt/hostedtoolcache/Python/3.12.4/x64/lib/python3.12/site-packages/nilearn/datasets/data/fsaverage5/infl_left.gii.gz
Fsaverage5 sulcal depth map of left hemisphere is at: /opt/hostedtoolcache/Python/3.12.4/x64/lib/python3.12/site-packages/nilearn/datasets/data/fsaverage5/sulc_left.gii.gz
Fsaverage5 curvature map of left hemisphere is at: /opt/hostedtoolcache/Python/3.12.4/x64/lib/python3.12/site-packages/nilearn/datasets/data/fsaverage5/curv_left.gii.gz

Extracting the seed time series

# Load resting state time series from nilearn
from nilearn import surface

timeseries = surface.load_surf_data(nki_dataset["func_left"][0])

# Coercing to float is required to avoid errors withj scipy >= 0.14.0
timeseries = timeseries.astype(float)

# Extract seed region via label
pcc_region = b"G_cingul-Post-dorsal"

import numpy as np

pcc_labels = np.where(parcellation == labels.index(pcc_region))[0]

# Extract time series from seed region
seed_timeseries = np.mean(timeseries[pcc_labels], axis=0)

Calculating seed-based functional connectivity

# Calculate Pearson product-moment correlation coefficient between seed
# time series and timeseries of all cortical nodes of the hemisphere
from scipy import stats

stat_map = np.zeros(timeseries.shape[0])
for i in range(timeseries.shape[0]):
    stat_map[i] = stats.pearsonr(seed_timeseries, timeseries[i])[0]

# Re-mask previously masked nodes (medial wall)
stat_map[np.where(np.mean(timeseries, axis=1) == 0)] = 0

/home/runner/work/nilearn/nilearn/examples/01_plotting/plot_surf_stat_map.py:110: ConstantInputWarning:

An input array is constant; the correlation coefficient is not defined.

Display ROI on surface

# Transform ROI indices in ROI map
pcc_map = np.zeros(parcellation.shape[0], dtype=int)
pcc_map[pcc_labels] = 1

from nilearn import plotting

plotting.plot_surf_roi(
    fsaverage["pial_left"],
    roi_map=pcc_map,
    hemi="left",
    view="medial",
    bg_map=fsaverage["sulc_left"],
    bg_on_data=True,
    title="PCC Seed",
)

<Figure size 400x500 with 1 Axes>

Using a flat mesh can be useful in order to easily locate the area of interest on the cortex. To make this plot easier to read, we use the mesh curvature as a background map.

bg_map = np.sign(surface.load_surf_data(fsaverage["curv_left"]))
# np.sign yields values in [-1, 1]. We rescale the background map
# such that values are in [0.25, 0.75], resulting in a nicer looking plot.
bg_map_rescaled = (bg_map + 1) / 4 + 0.25

plotting.plot_surf_roi(
    fsaverage["flat_left"],
    roi_map=pcc_map,
    hemi="left",
    view="dorsal",
    bg_map=bg_map_rescaled,
    bg_on_data=True,
    title="PCC Seed",
)

<Figure size 400x500 with 1 Axes>

Display unthresholded stat map with a slightly dimmed background

plotting.plot_surf_stat_map(
    fsaverage["pial_left"],
    stat_map=stat_map,
    hemi="left",
    view="medial",
    colorbar=True,
    bg_map=fsaverage["sulc_left"],
    bg_on_data=True,
    darkness=0.3,
    title="Correlation map",
)

<Figure size 470x500 with 2 Axes>

Many different options are available for plotting, for example thresholding, or using custom colormaps

plotting.plot_surf_stat_map(
    fsaverage["pial_left"],
    stat_map=stat_map,
    hemi="left",
    view="medial",
    colorbar=True,
    bg_map=fsaverage["sulc_left"],
    bg_on_data=True,
    cmap="Spectral",
    threshold=0.5,
    title="Threshold and colormap",
)

<Figure size 470x500 with 2 Axes>

Here the surface is plotted in a lateral view without a background map. To capture 3D structure without depth information, the default is to plot a half transparent surface. Note that you can also control the transparency with a background map using the alpha parameter.

plotting.plot_surf_stat_map(
    fsaverage["pial_left"],
    stat_map=stat_map,
    hemi="left",
    view="lateral",
    colorbar=True,
    cmap="Spectral",
    threshold=0.5,
    title="Plotting without background",
)

<Figure size 470x500 with 2 Axes>

The plots can be saved to file, in which case the display is closed after creating the figure

from pathlib import Path

output_dir = Path.cwd() / "results" / "plot_surf_stat_map"
output_dir.mkdir(exist_ok=True, parents=True)
print(f"Output will be saved to: {output_dir}")

plotting.plot_surf_stat_map(
    fsaverage["infl_left"],
    stat_map=stat_map,
    hemi="left",
    bg_map=fsaverage["sulc_left"],
    bg_on_data=True,
    threshold=0.5,
    colorbar=True,
    output_file=output_dir / "plot_surf_stat_map.png",
)

plotting.show()

Output will be saved to: /home/runner/work/nilearn/nilearn/examples/01_plotting/results/plot_surf_stat_map

References

[1]

Kate Nooner, Stanley Colcombe, Russell Tobe, Maarten Mennes, Melissa Benedict, Alexis Moreno, Laura Panek, Shaquanna Brown, Stephen Zavitz, Qingyang Li, Sharad Sikka, David Gutman, Saroja Bangaru, Rochelle Tziona Schlachter, Stephanie Kamiel, Ayesha Anwar, Caitlin Hinz, Michelle Kaplan, Anna Rachlin, Samantha Adelsberg, Brian Cheung, Ranjit Khanuja, Chaogan Yan, Cameron Craddock, Vincent Calhoun, William Courtney, Margaret King, Dylan Wood, Christine Cox, Clare Kelly, Adriana DiMartino, Eva Petkova, Philip Reiss, Nancy Duan, Dawn Thompsen, Bharat Biswal, Barbara Coffey, Matthew Hoptman, Daniel Javitt, Nunzio Pomara, John Sidtis, Harold Koplewicz, Francisco Castellanos, Bennett Leventhal, and Michael Milham. The nki-rockland sample: a model for accelerating the pace of discovery science in psychiatry. Frontiers in Neuroscience, 6:152, 2012. URL: https://www.frontiersin.org/article/10.3389/fnins.2012.00152, doi:10.3389/fnins.2012.00152.

[2]

Anders M. Dale, Bruce Fischl, and Martin I. Sereno. Cortical surface-based analysis: i. segmentation and surface reconstruction. NeuroImage, 9(2):179–194, 1999. URL: https://www.sciencedirect.com/science/article/pii/S1053811998903950, doi:https://doi.org/10.1006/nimg.1998.0395.

[3]

Bruce Fischl, Martin I. Sereno, and Anders M. Dale. Cortical surface-based analysis: ii: inflation, flattening, and a surface-based coordinate system. NeuroImage, 9(2):195–207, 1999. URL: https://www.sciencedirect.com/science/article/pii/S1053811998903962, doi:https://doi.org/10.1006/nimg.1998.0396.

[4]

Christophe Destrieux, Bruce Fischl, Anders Dale, and Eric Halgren. Automatic parcellation of human cortical gyri and sulci using standard anatomical nomenclature. NeuroImage, 53(1):1–15, 2010. URL: https://www.sciencedirect.com/science/article/pii/S1053811910008542, doi:https://doi.org/10.1016/j.neuroimage.2010.06.010.

Estimated memory usage: 537 MB