Light on broken networks: Resting-state fNIRS as a tool for connectivity mapping

Area of Science:

• Neuroscience
• Brain Imaging
• Systems Neuroscience

Background:

• Resting-state functional connectivity (RSFC) and networks (RSNs) are crucial for understanding brain organization and neurological disease.
• Functional MRI (fMRI) is standard for RSN assessment but limited by cost, motion sensitivity, and feasibility for repeated measures.
• Functional near-infrared spectroscopy (fNIRS) offers a portable alternative, but its reliability for RSFC and RSN mapping requires further validation.

Purpose of the Study:

• To compare the reliability and cross-modal correspondence of RSNs derived from fNIRS and fMRI.
• To assess the utility of fNIRS for capturing large-scale brain organization comparable to fMRI.
• To evaluate the impact of correlation methods (bivariate vs. partial) on cross-modal agreement at different analytical levels.

Main Methods:

• Acquired near whole-head fNIRS data and fMRI-BOLD signals from corresponding cortical regions in two independent cohorts (n=31 each).
• Compared RSN organization using bivariate and partial correlations across edge, nodal, and network levels.
• Assessed cross-modal convergence and divergence using graph theory metrics and module overlap analysis (Jaccard index).

Main Results:

• Substantial modality differences in edge-wise connectivity (50-61%) reduced significantly with partial correlations (<3%).
• Moderate cross-modal similarity in group-averaged connectivity patterns (r ≈ 0.37) and broadly similar graph-metric distributions.
• fNIRS-derived modules significantly overlapped with fMRI modules, identifying major RSNs (Jaccard ≈ 0.27-0.5).

Conclusions:

• fNIRS effectively captures key features of large-scale RSFC and RSN organization comparable to fMRI.
• Cross-modal correspondence supports the translational utility of fNIRS for brain network research.
• Partial correlations improve edge-level agreement but may attenuate nodal/modular recovery, favoring targeted analyses over whole-network characterization.