Multiscale Functional Connectivity: Exploring the brain functional connectivity at different timescales

TL;DR

This paper introduces a novel multiscale functional connectivity method for fMRI data that decomposes signals into intrinsic modes, capturing brain dynamics across multiple timescales while accounting for nonlinear and nonstationary properties.

Contribution

The study presents a new multiscale analysis approach for fMRI data that improves the separation of neurophysiological signals at different timescales and enhances the understanding of brain connectivity.

Findings

Effective separation of fMRI data into multiple timescales.

Identification of reliable functional connectivity patterns.

Insights into how FC spans different timescales across experiments.

Abstract

Human brains exhibit highly organized multiscale neurophysiological dynamics. Understanding those dynamic changes and the neuronal networks involved is critical for understanding how the brain functions in health and disease. Functional Magnetic Resonance Imaging (fMRI) is a prevalent neuroimaging technique for studying these complex interactions. However, analyzing fMRI data poses several challenges. Furthermore, most approaches for analyzing Functional Connectivity (FC) still rely on preprocessing or conventional methods, often built upon oversimplified assumptions. On top of that, those approaches often ignore frequency-related information despite evidence showing that fMRI data contain rich information that spans multiple timescales. This study introduces a novel methodology, Multiscale Functional Connectivity (MFC), to analyze fMRI data by decomposing the fMRI into their intrinsic modes, allowing us to separate the neurophysiological activation patterns at multiple timescales while separating them from other interfering components. Additionally, the proposed approach accounts for the natural nonlinear and nonstationary nature of fMRI and the particularities of each individual in a data-driven way. We evaluated the performance of our proposed methodology using three fMRI experiments. Our results demonstrate that our novel approach effectively separates the fMRI data into different timescales while identifying highly reliable functional connectivity patterns across individuals. In addition, we further extended our knowledge of how the FC for these three experiments spans among different timescales.