Statistical Analysis of Dynamic Functional Brain Networks in Twins

TL;DR

This paper introduces a heat kernel-based method for more reliable estimation of dynamic functional brain networks from fMRI data, demonstrating improved accuracy and heritability analysis in twin studies.

Contribution

The paper presents a novel heat kernel approach to enhance the robustness of dynamic correlation estimates in brain networks, addressing noise and high-frequency fluctuations.

Findings

Heat kernel method reduces high-frequency noise in correlation estimates.

Improved detection of dynamic brain states compared to traditional methods.

Evidence of genetic heritability in dynamic brain network changes.

Abstract

Recent studies have shown that functional brain brainwork is dynamic even during rest. A common approach to modeling the brain network in whole brain resting-state fMRI is to compute the correlation between anatomical regions via sliding windows. However, the direct use of the sample correlation matrices is not reliable due to the image acquisition, processing noises and the use of discrete windows that often introduce spurious high-frequency fluctuations and the zig-zag pattern in the estimated time-varying correlation measures. To address the problem and obtain more robust correlation estimates, we propose the heat kernel based dynamic correlations. We demonstrate that the proposed heat kernel method can smooth out the unwanted high-frequency fluctuations in correlation estimations and achieve higher accuracy in identifying dynamically changing distinct states. The method is further used in determining if such dynamic state change is genetically heritable using a large-scale twin study. Various methodological challenges for analyzing paired twin dynamic networks are addressed.