Robust dynamic community detection with applications to human brain functional networks

L.-E Martinet,E D Kolaczyk,S S Cash,L N Perkins,W Viles,C J Chu,E Spencer,M A Kramer

doi:10.1038/s41467-020-16285-7

L.-E Martinet, E D Kolaczyk + Show 6 more

Open Access

https://doi.org/10.1038/s41467-020-16285-7

Copy DOI

Abstract

While current technology permits inference of dynamic brain networks over long time periods at high temporal resolution, the detailed structure of dynamic network communities during human seizures remains poorly understood. We introduce a new methodology that addresses critical aspects unique to the analysis of dynamic functional networks inferred from noisy data. We propose a dynamic plex percolation method (DPPM) that is robust to edge noise, and yields well-defined spatiotemporal communities that span forward and backwards in time. We show in simulation that DPPM outperforms existing methods in accurately capturing certain stereotypical dynamic community behaviors in noisy situations. We then illustrate the ability of this method to track dynamic community organization during human seizures, using invasive brain voltage recordings at seizure onset. We conjecture that application of this method will yield new targets for surgical treatment of epilepsy, and more generally could provide new insights in other network neuroscience applications.

Highlights

While current technology permits inference of dynamic brain networks over long time periods at high temporal resolution, the detailed structure of dynamic network communities during human seizures remains poorly understood
The dynamic plex percolation method (DPPM) operates on dynamic networks, which may be inferred from noisy time series data
We define a community explicitly as subsets of nodes—within and across time—that are reachable by small template subgraphs that are walked within and across temporally adjacent graphs Gt. The result of this definition is that our communities may be conceptualized as an evolving series of tubes, which represent cohesive aspects of the dynamic networks evolving in time

Summary

Introduction

While current technology permits inference of dynamic brain networks over long time periods at high temporal resolution, the detailed structure of dynamic network communities during human seizures remains poorly understood. Many tools have emerged to characterize network dynamics[16,17], the most common is dynamic community detection (i.e., tracking how a group of nodes that share increased connections changes in time). These methods typically apply an algorithm developed for static graphs to define candidate communities at a fixed time, and define time-varying communities by linking consecutive static communities. Existing methods generally are unable to account for the edge noise that is inherently present in functional networks inferred from noisy brain voltage data[28]. MMM employs an optimization criterion[19,30], yielding an implicit notion of community (albeit computationally tractable and mathematically elegant)

Objectives

Methods

Results

Conclusion