Resting-State Functional
Connectivity Matrices

CC200

Generated via spatially-constrained normalized-cut spectral clustering on 41 healthy controls from the 1000FCP dataset itself. Grey matter voxels clustered into ~200 spatially contiguous, functionally homogeneous regions. Recommended for most connectivity studies — good balance of interpretability and functional accuracy.

CC400

Same clustering procedure as CC200 but at higher spatial resolution (~400 parcels). Provides finer-grained connectivity estimates; preferred when FC accuracy is prioritized over anatomical interpretability. Higher ROI count reduces signal-to-noise ratio per ROI.

AAL

The classic Tzourio-Mazoyer 2002 automated anatomical labeling atlas. Parcels defined by macro-anatomical landmarks (gyral boundaries). Widely used for cross-study comparability. 90 cortical + 26 cerebellar regions. Known to have poor functional homogeneity within large anatomical regions.

Dosenbach264

Meta-analytically defined set of 264 putative functional areas, derived from coordinate-based meta-analysis of task fMRI studies. Each ROI is a 5mm-radius sphere centered on a peak activation coordinate. Covers cortex and subcortex. Widely used in graph-theoretic network analysis of the functional connectome.

openconnectome/1000FCP/ ├── matrices/ │ ├── CC200/ # 200-ROI Craddock parcellation │ │ ├── <site>_<subject>_CC200.mat │ │ └── ... │ ├── CC400/ # 400-ROI Craddock parcellation │ ├── AAL/ # 116-ROI AAL atlas │ └── Dosenbach264/ # 264-ROI Power/Dosenbach atlas ├── masks/ │ ├── <site>_grey_matter_mask.nii.gz │ └── ... └── README.md

Matrix Files (.mat)

• MATLAB .mat format (v7.3 compatible)
• Variable name: connectivity
• Shape: N×N (symmetric)
• Values: Pearson r (or Fisher-z)
• Diagonal: NaN or 1.0 depending on version
• Readable by Python scipy.io.loadmat

Mask Files (.nii.gz)

• NIfTI-1 gzipped format
• Binary grey matter masks per site
• MNI152 standard space (2mm isotropic)
• Used to define valid voxels for parcellation
• Derived from FreeSurfer tissue segmentation

import scipy.io import numpy as np import glob, os # Load all CC200 matrices for one site site = 'Beijing_Zang' files = glob.glob(f'matrices/CC200/{site}_*.mat') matrices = [] for f in files: data = scipy.io.loadmat(f) conn = data['connectivity'] # 200×200 np.fill_diagonal(conn, 0) # zero diagonal matrices.append(conn) # Stack into 3D array (n_subjects × 200 × 200) matrices = np.stack(matrices) # Compute group-average connectivity matrix group_mean = np.mean(matrices, axis=0) # Fisher-z transform for averaging across subjects z_matrices = np.arctanh(np.clip(matrices, -0.9999, 0.9999)) group_z = np.mean(z_matrices, axis=0) group_r = np.tanh(group_z) # back-transform # Graph-theoretic measures (requires networkx) import networkx as nx threshold = 0.2 # proportional threshold adj = (group_r > threshold).astype(float) G = nx.from_numpy_array(adj) print(f"Global efficiency: {nx.global_efficiency(G):.4f}") print(f"Clustering coeff: {nx.average_clustering(G):.4f}")

% Load a single subject's CC200 RSFC matrix data = load('matrices/CC200/Beijing_Zang_sub001_CC200.mat'); conn = data.connectivity; % 200×200 Pearson r matrix % Visualize as heatmap figure; imagesc(conn, [-1 1]); colormap('jet'); colorbar; title('1000FCP CC200 Connectivity Matrix'); xlabel('ROI'); ylabel('ROI'); % Group-level analysis: load all subjects, average with Fisher-z files = dir('matrices/CC200/Beijing_Zang_*.mat'); all_z = zeros(200, 200, length(files)); for i = 1:length(files) d = load(fullfile(files(i).folder, files(i).name)); all_z(:,:,i) = atanh(d.connectivity); end group_r = tanh(mean(all_z, 3));