In the absence of a standard for brain parcellations, the following minimum requirements are recommended:

• Ensure comprehensive coverage of functional regions throughout the brain, including cortical, subcortical structures, and the cerebellum.

• Divide the brain into at least 200 functional regions.

• Base the delineation of regions on FC, potentially combined with multimodal imaging.

• Ensure regions exhibit high FC homogeneity.

• Provide clear guidance on the modular structure of regions within the parcellation, discouraging researchers from identifying functional modules when a published modular structure is available.

• Longer resting-state acquisitions improve the stability and test-retest reliability of FC estimates.

• Clinical researchers should collect resting-state data for at least 9 min.

• Faster temporal sampling (e.g., 1-second TR) should be used, when possible, potentially utilizing multiband imaging.

• The reliability of FC based on partial correlation decreases with an increasing number of nodes or fewer measurements, requiring longer scan times.

• Bivariate correlations are more stable, typically reaching consistency after 250 measurements, regardless of node count.

• For conventional resting-state data (e.g., 180 volumes, 2-second TR), using a shrinkage estimator of marginal correlation is recommended over conditional association measures.

• Proportional thresholding should be avoided in case-control studies as it can obscure or distort results.

• Researchers should clearly describe how negative FC estimates are handled, as deleting negative edges is an untested assumption. If many negative edges are present, they should be reported and possibly analyzed using a separate graph.

To promote formal comparisons across studies, researchers are encouraged to report a standard set of graph metrics. A minimal set includes:

• Global clustering coefficient.

• Average path length.

• Modularity.

• Degree.

• Eigenvector centrality.

• Summary statistics of edge strength.

• Researchers should conceptualize and report graph analyses across different levels of analysis, from global to specific.

• Global metrics may overlook regional effects of pathology or neurodevelopment, so careful consideration is needed.

• Researchers should ensure alignment between the graph analysis level and the biological understanding of neuropathology.

• The relevance of certain graph metrics, like small-worldness, to understanding brain disorders is still unclear [77].