Telling functional networks apart using ranked network features stability

Massimiliano Zanin,Bahar Güntekin,Ebru Yıldırım,Tuba Aktürk,Ilayda Kiyi,Görsev Yener,Duygu Hünerli-Gündüz,Henrique Sequeira,David Papo

doi:10.1038/s41598-022-06497-w

Abstract

Over the past few years, it has become standard to describe brain anatomical and functional organisation in terms of complex networks, wherein single brain regions or modules and their connections are respectively identified with network nodes and the links connecting them. Often, the goal of a given study is not that of modelling brain activity but, more basically, to discriminate between experimental conditions or populations, thus to find a way to compute differences between them. This in turn involves two important aspects: defining discriminative features and quantifying differences between them. Here we show that the ranked dynamical stability of network features, from links or nodes to higher-level network properties, discriminates well between healthy brain activity and various pathological conditions. These easily computable properties, which constitute local but topographically aspecific aspects of brain activity, greatly simplify inter-network comparisons and spare the need for network pruning. Our results are discussed in terms of microstate stability. Some implications for functional brain activity are discussed.

Highlights

Over the past few years, it has become standard to describe brain anatomical and functional organisation in terms of complex networks, wherein single brain regions or modules and their connections are respectively identified with network nodes and the links connecting them
The global organisation of healthy brain activity can be thought of as a balance between these two modes[9–11], and imbalance between segregation and integration has been associated with various pathological conditions, e.g., autism or schizophrenia10–12, epilepsy[13,14], and LSD consumption[15]
This organisation is naturally described in terms of complex networks[16,17], wherein single brain regions or modules and their connections are respectively identified with network nodes and the links connecting t hem[18]

Summary

Introduction

Over the past few years, it has become standard to describe brain anatomical and functional organisation in terms of complex networks, wherein single brain regions or modules and their connections are respectively identified with network nodes and the links connecting them. The goal of a given study is not that of modelling brain activity but, more basically, to discriminate between experimental conditions or populations, to find a way to compute differences between them This in turn involves two important aspects: defining discriminative features and quantifying differences between them. The global organisation of healthy brain activity can be thought of as a balance between these two modes[9–11], and imbalance between segregation and integration has been associated with various pathological conditions, e.g., autism or schizophrenia10–12, epilepsy[13,14], and LSD consumption[15] This organisation is naturally described in terms of complex networks[16,17], wherein single brain regions or modules and their connections are respectively identified with network nodes and the links connecting t hem[18]. Weak links may be key to brain function[27], and to the discrimination between brain pathologies[21], and retaining higher percentages of links may improve classification a ccuracy[28]

Methods

Results

Conclusion