Massively univariate analysis of a motor task from the Localizer dataset

This example shows the results obtained in a massively univariate analysis performed at the inter-subject level with various methods. We use the [left button press (auditory cue)] task from the Localizer dataset and seek association between the contrast values and a variate that measures the speed of pseudo-word reading. No confounding variate is included in the model.

  1. A standard ANOVA is performed. Data smoothed at 5 voxels FWHM are used.

  2. A permuted Ordinary Least Squares algorithm is run at each voxel. Data smoothed at 5 voxels FWHM are used.

Note

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
from nilearn._utils.helpers import check_matplotlib

check_matplotlib()
import numpy as np

from nilearn import datasets
from nilearn.maskers import NiftiMasker
from nilearn.mass_univariate import permuted_ols

Load Localizer contrast

n_samples = 94
localizer_dataset = datasets.fetch_localizer_contrasts(
    ["left button press (auditory cue)"],
    n_subjects=n_samples,
    legacy_format=False,
)

# print basic information on the dataset
print(
    "First contrast nifti image (3D) is located "
    f"at: {localizer_dataset.cmaps[0]}"
)

tested_var = localizer_dataset.ext_vars["pseudo"]

# Quality check / Remove subjects with bad tested variate
mask_quality_check = np.where(np.logical_not(np.isnan(tested_var)))[0]
n_samples = mask_quality_check.size
contrast_map_filenames = [
    localizer_dataset.cmaps[i] for i in mask_quality_check
]
tested_var = tested_var[mask_quality_check].to_numpy().reshape((-1, 1))
print(f"Actual number of subjects after quality check: {int(n_samples)}")
[get_dataset_dir] Dataset found in
/home/runner/nilearn_data/brainomics_localizer
First contrast nifti image (3D) is located at: /home/runner/nilearn_data/brainomics_localizer/brainomics_data/S01/cmaps_LeftAuditoryClick.nii.gz
Actual number of subjects after quality check: 89

Mask data

nifti_masker = NiftiMasker(
    smoothing_fwhm=5, memory="nilearn_cache", memory_level=1
)
fmri_masked = nifti_masker.fit_transform(contrast_map_filenames)

Anova (parametric F-scores)

/home/runner/work/nilearn/nilearn/.tox/doc/lib/python3.9/site-packages/sklearn/utils/validation.py:1229: DataConversionWarning:

A column-vector y was passed when a 1d array was expected. Please change the shape of y to (n_samples, ), for example using ravel().

Perform massively univariate analysis with permuted OLS

This method will produce both voxel-level FWE-corrected -log10 p-values and TFCE-based FWE-corrected -log10 p-values.

Note

permuted_ols can support a wide range of analysis designs, depending on the tested_var. For example, if you wished to perform a one-sample test, you could simply provide an array of ones (e.g., np.ones(n_samples)).

ols_outputs = permuted_ols(
    tested_var,  # this is equivalent to the design matrix, in array form
    fmri_masked,
    model_intercept=True,
    masker=nifti_masker,
    tfce=True,
    n_perm=200,  # 200 for the sake of time. Ideally, this should be 10000.
    verbose=1,  # display progress bar
    n_jobs=2,  # can be changed to use more CPUs
    output_type="dict",
)
neg_log_pvals_permuted_ols_unmasked = nifti_masker.inverse_transform(
    ols_outputs["logp_max_t"][0, :]  # select first regressor
)
neg_log_pvals_tfce_unmasked = nifti_masker.inverse_transform(
    ols_outputs["logp_max_tfce"][0, :]  # select first regressor
)
[Parallel(n_jobs=2)]: Using backend LokyBackend with 2 concurrent workers.
[Parallel(n_jobs=2)]: Done   2 out of   2 | elapsed:   49.9s remaining:    0.0s
[Parallel(n_jobs=2)]: Done   2 out of   2 | elapsed:   49.9s finished

Visualization

import matplotlib.pyplot as plt

from nilearn import plotting
from nilearn.image import get_data

# Various plotting parameters
z_slice = 12  # plotted slice
threshold = -np.log10(0.1)  # 10% corrected

vmax = max(
    np.amax(ols_outputs["logp_max_t"]),
    np.amax(neg_log_pvals_anova),
    np.amax(ols_outputs["logp_max_tfce"]),
)

images_to_plot = {
    "Parametric Test\n(Bonferroni FWE)": neg_log_pvals_anova_unmasked,
    "Permutation Test\n(Max t-statistic FWE)": (
        neg_log_pvals_permuted_ols_unmasked
    ),
    "Permutation Test\n(Max TFCE FWE)": neg_log_pvals_tfce_unmasked,
}

fig, axes = plt.subplots(figsize=(12, 3), ncols=3)
for i_col, (title, img) in enumerate(images_to_plot.items()):
    ax = axes[i_col]
    n_detections = (get_data(img) > threshold).sum()
    new_title = f"{title}\n{n_detections} sig. voxels"

    plotting.plot_glass_brain(
        img,
        colorbar=True,
        vmax=vmax,
        display_mode="z",
        plot_abs=False,
        cut_coords=[12],
        threshold=threshold,
        figure=fig,
        axes=ax,
    )
    ax.set_title(new_title)

fig.suptitle(
    "Group left button press ($-\\log_{10}$ p-values)",
    y=1.3,
    fontsize=16,
)
Group left button press ($-\log_{10}$ p-values), Parametric Test (Bonferroni FWE) 3 sig. voxels, Permutation Test (Max t-statistic FWE) 13 sig. voxels, Permutation Test (Max TFCE FWE) 675 sig. voxels
Text(0.5, 1.3, 'Group left button press ($-\\log_{10}$ p-values)')

Total running time of the script: (0 minutes 58.534 seconds)

Estimated memory usage: 303 MB

Gallery generated by Sphinx-Gallery