Meta-analytic coactivation modeling analysis

Perform a MACM analysis with Neurosynth data.

Meta-analytic coactivation modeling (MACM) is a common coordinate-based analysis in which task-independent “connectivity” is assessed by selecting studies within a larger database based on locations of report coordinates.

import nibabel as nib
import numpy as np
from nilearn import datasets, image, plotting

from nimare.correct import FWECorrector
from nimare.dataset import Dataset
from nimare.meta.cbma.ale import SCALE
from nimare.meta.cbma.mkda import MKDAChi2

Load Dataset

We will assume that the Neurosynth database has already been downloaded and converted to a NiMARE dataset.

dset_file = "neurosynth_nimare_with_abstracts.pkl.gz"
dset = Dataset.load(dset_file)

Define a region of interest

We’ll use the right amygdala from the Harvard-Oxford atlas

atlas = datasets.fetch_atlas_harvard_oxford("sub-maxprob-thr50-2mm")
img = atlas["maps"]

roi_idx = atlas["labels"].index("Right Amygdala")
img_vals = np.unique(img.get_fdata())
roi_val = img_vals[roi_idx]
roi_img = image.math_img(f"img1 == {roi_val}", img1=img)

Select studies with a reported coordinate in the ROI

roi_ids = dset.get_studies_by_mask(roi_img)
dset_sel = dset.slice(roi_ids)
print(f"{len(roi_ids)}/{len(dset.ids)} studies report at least one coordinate in the ROI")

Select studies with no reported coordinates in the ROI

no_roi_ids = list(set(dset.ids).difference(roi_ids))
dset_unsel = dset.slice(no_roi_ids)
print(f"{len(no_roi_ids)}/{len(dset.ids)} studies report zero coordinates in the ROI")

MKDA Chi2 with FWE correction

mkda = MKDAChi2(kernel__r=10)
results = mkda.fit(dset_sel, dset_unsel)

corr = FWECorrector(method="montecarlo", n_iters=10000)
cres = corr.transform(results)

# We want the "association" map (2-way chi-square between sel and unsel)
plotting.plot_stat_map(
    cres.get_map("z_desc-uniformity_level-voxel_corr-FWE_method-montecarlo"),
    threshold=3.09,
    draw_cross=False,
    cmap="RdBu_r",
)

SCALE

Another good option for a MACM analysis is the SCALE algorithm, which was designed specifically for MACM. Unfortunately, SCALE does not support multiple-comparisons correction.

# First, we must define our null model of reported coordinates in the literature.
# We will use the coordinates in Neurosynth
xyz = dset.coordinates[["x", "y", "z"]].values
scale = SCALE(xyz=xyz, n_iters=10000, n_cores=1, kernel__n=20)
results = scale.fit(dset_sel)
plotting.plot_stat_map(results.get_map("z"), draw_cross=False, cmap="RdBu_r")