Review Papers Claim by Claim

 Quick Explanation Copied

 Long Explanation

Evidence Basis for the Central Claim

The study compared 82-channel fNIRS (HbO/HbR) against HCP rs-fMRI in two independent healthy cohorts (N = 31 each, matched for age/sex) across four analytical scales Primary fNIRS-fMRI connectivity paper.

Group-averaged connectivity matrices correlated moderately across modalities (HbO: r = 0.393; HbR: r = 0.353; partial correlations: 0.379–0.385; all P < 10⁻⁵, permutation-tested). Subject-level edgewise t-tests revealed large modality differences (50.3% of edges for HbO, 61.4% for HbR with bivariate correlations), collapsing to <3% with partial correlations — the authors correctly note this may partly reflect sparsification rather than true convergence Edgewise t-test results.

Nodal & Modular Findings

Normalized nodal strength (rho = 0.51 HbO) and assortativity showed strongest cross-modal agreement; local efficiency and path length diverged markedly, especially under partial correlations (100% of nodes differing) Nodal graph-metric comparison. Louvain modules overlapping significantly with canonical Yeo networks (DMN Jaccard ≈ 0.47, CEN/SN 0.50) but mean overlap only ~0.31–0.36; fMRI modularity (Q ≈ 0.40–0.52) exceeded fNIRS (Q ≈ 0.15–0.36), consistent with lower fNIRS spatial resolution and coverage Modularity and Jaccard results.

Critical Appraisal

• Cohort confound: modalities were not simultaneously recorded and used different participants (HCP vs local fNIRS), so cross-modal r ≈ 0.35–0.39 confounds modality with cohort/idiosyncratic variability. The authors acknowledge this and argue correspondence could be stronger under matched conditions, but this remains untested inference.
• Duration mismatch: 5-min fNIRS vs 15-min fMRI reduces fNIRS RSFC reliability — a known driver of attenuated cross-modal correlation.
• Mapping loss: 20 of 102 channels excluded due to fOLD coordinate failures, degrading motor network coverage; 0.5 cm spherical ROIs introduce spatial uncertainty.
• Blindspots: healthy young adults only (mean 28 y); clinical extrapolation is speculative; betweenness centrality unassessed; no simultaneous within-subject validation — the definitive test remains absent.
• Equivalence testing (TOST) is a methodological strength rarely applied in this literature, giving interpretable equivalence bounds (Δr ≈ 0.11 partial; ≈ 0.21–0.25 bivariate).

Conclusion & Falsification

With moderate confidence, the evidence supports group-level fNIRS–fMRI correspondence for large-scale RSN organization but not fine-grained or subject-specific connectivity. The conclusion would be overturned if simultaneous within-subject fNIRS-fMRI recordings failed to show above-chance group-level matrix correlation, or if module overlaps did not exceed permutation nulls in larger, independent cohorts. The paper does not overstate — it explicitly restricts translational claims to population-level monitoring.

Author Review: Foivos KotsogiannisAuthor Review: Michael LührsAuthor Review: Geert-Jan M. RuttenAuthor Review: Andrew T. ReidAuthor Review: Sabine DeprezAuthor Review: Maarten LambrechtAuthor Review: Wouter De BaeneAuthor Review: Charlotte Sleurs

 BGPT Paper Review

Study Novelty

First systematic multi-scale (edge/node/graph/module) fNIRS-fMRI RSFC comparison with near whole-head coverage and TOST equivalence testing; prior work covered limited regions or single scales.

Scientific Quality

Rigorous statistics (permutation tests, TOST, FDR, consensus Louvain), transparent limitations, no COI. However, cross-cohort design confounds modality with cohort effects; 5 vs 15-min duration mismatch; 20/102 channels lost to mapping; findings under- rather than over-claimed.

Study Generality

Generalizable framework for cross-modal connectivity validation, but conclusions limited to healthy adults and superficial cortical networks.

Study Usefulness

Directly informs which fNIRS metrics/analyses are trustworthy for clinical group-level RSN monitoring — a practical translational decision tool.

Study Reproducibility

Detailed methods (hardware, pipeline, parameters, Appendix A), public HCP comparison data; but fNIRS dataset not yet released, limiting full replication.

Explanatory Depth

Mechanistic reasoning about why partial correlations alter cross-modal agreement (sparsity, shared-variance removal), though causal explanation of modality differences is incomplete.