Work

Structural coarse-graining enables statistically controlled functional connectivity and reshapes inferred intersubject variability

Year: 2026
Type: article
Abstract: Functional connectivity (FC) estimates are highly sensitive to analysis choices and can be dominated by noise when the number of sampled time points is small relative to network dimensionality. This issue is particularly acute in fMRI, where scan duration is constrained by practical factors such as motion and fatigue, leaving many datasets statistically underpowered for high-dimensional correlation estimation. We introduce a framework that combines diffusion-based structural coarse-graining with spectral noise filtering to estimate functional networks at a statistically controlled resolution. The method groups regions according to diffusion-defined communication, producing coarse-grained networks whose size is compatible with the available time points, and then applies random matrix filtering to derivative-based correlations, which are insensitive to the slow non-stationary fluctuations that dominate raw-signal correlations. Synthetic benchmarks show that this procedure retains network-shaped collective correlations even when time series are shorter than the number of nodes. In resting-state fMRI data, the choice of FC estimator changes which subjects are grouped together; after filtering, the population separates into functional organization modes that differ mainly in the strength of the collective component of connectivity and, correspondingly, of inter-modular anticorrelations, differences that are already present, though less pronounced, in raw correlations. These results provide a practical route to functional networks under realistic sampling constraints and make explicit which collective modes can be distinguished from sampling noise at a given resolution.
Language: English
Open Access status: gold
Estimated APC paid: $2,625
OpenAlex ID: W7128497716
FWCI: 1.705
Cites: 50
Cited by: 1
Related to: 10