End-to-end pipeline for multi-atlas coordinate querying

Generates random MNI coordinates, retrieves regions from several atlases, and queries related studies, saving results to disk.

import os
import json
import numpy as np
import pandas as pd
from coord2region.fetching import AtlasFetcher
from coord2region.coord2region import MultiAtlasMapper
from coord2region.coord2study import prepare_datasets, get_studies_for_coordinate

Step 1: Generate 10 random MNI coordinates

np.random.seed(42)
mni_coordinates = np.random.randint(-60, 60, size=(10, 3))  # Generate 10 random (x, y, z) points
mni_list = mni_coordinates.tolist()
# Save coordinates to a CSV file
csv_filename = "toy_coordinates.csv"
df = pd.DataFrame(mni_list, columns=["x", "y", "z"])
df.to_csv(csv_filename, index=False)
print(f"Saved toy coordinates to {csv_filename}")
# df = pd.read_csv("coordinates.csv"); instead of generating random coordinates

Step 2: Fetch and load multiple atlases

af = AtlasFetcher()
atlases = {
    "harvard-oxford": {},
    "schaefer": {},
    "yeo": {},
    "juelich": {},
}

Step 3: Create MultiAtlasMapper

multi_mapper = MultiAtlasMapper(data_dir=af.data_dir, atlases=atlases)

Step 4: Map each coordinate to regions in all four atlases

results = []
for coord in mni_list:
    regions = multi_mapper.batch_mni_to_region_names([coord])
    result = {"x": coord[0], "y": coord[1], "z": coord[2]}
    for atlas, region_list in regions.items():
        result[atlas] = region_list[0] if region_list else "Unknown"
    results.append(result)
# Save results to a CSV file
mapped_csv_filename = "mapped_coordinates.csv"
df_mapped = pd.DataFrame(results)
df_mapped.to_csv(mapped_csv_filename, index=False)
print(f"Saved mapped coordinates to {mapped_csv_filename}")

Step 5: Prepare NiMARE dataset (only NIDM-Pain for speed)

dataset_dir = af.data_dir
# Neurosynth and NeuroQuery take a long time to download; for this example we
# only include the NIDM-Pain dataset.
dataset = prepare_datasets(dataset_dir, sources=["nidm_pain"])
datasets = {"Combined": dataset}

Step 6: Query studies for each coordinate and save to JSON

output_dir = "studies_per_coordinate"
os.makedirs(output_dir, exist_ok=True)
for coord in mni_list:
    studies = get_studies_for_coordinate(datasets, coord)
    coord_filename = f"{output_dir}/studies_{coord[0]}_{coord[1]}_{coord[2]}.json"
    with open(coord_filename, "w") as f:
        json.dump(studies, f, indent=4)
    print(f"Saved studies for coordinate {coord} to {coord_filename}")

Summary:

• We generated 10 random MNI coordinates and saved them to toy_coordinates.csv.

• We mapped each coordinate to 4 brain atlases and saved the results to mapped_coordinates.csv.

• We prepared a deduplicated NIDM-Pain dataset and queried studies for each coordinate, saving them to JSON files.