Multifractal analysis of Resting State Networks in functional MRI

Abstract

It has been know for at least one decade [1] that functional MRI time series display long-memory properties, such as power-law scaling in the frequency spectrum. Concomitantly, multivariate modelfree analysis of spatial patterns , such as spatial Independent Component Analysis (sICA) [2], has been successfully used to segment from spontaneous activity Resting-State Networks (RSN) that correspond to known brain function. As recent neuroscientific studies suggest a link between spectral properties of brain activity and cognitive processes [3], a burning question emerges: can temporal scaling properties offer new markers of brain states encoded in these large scale networks? In this paper, we combine two recent methodologies: group-level canonical ICA for multi-subject segmentation of brain network, and wavelet leader-based multifractal formalism for the analysis of RSN scaling properties. We identify the brain networks that elicit self-similarity or multifractality and explore which spectral properties correspond specifically to known functionallyrelevant processes in spontaneous activity.

Modulation of scale-free properties of brain activity in MEG

“…it has been already used successfully to analyse multifractal properties of fMRI signals [5] and scale-free dynamics in EEG microstates [7]. In the following, it allows us to accurately characterize scale-free properties of MEG time series and their modulation by external stimulus.…”

“…However, several authors have shown that ongoing activity, the major part of brain activity [2,3], has scale-free dynamics -i.e. a 1/f power spectrum [1,[3][4][5]. Since then, the study of scale-free properties has emerged as a new research topic in neuroscience and brain neuroimaging.…”

“…This enables to correctly estimate scale invariance parameters even when data present various forms of non-stationarity. Moreover, some empirical results [5] suggest that ongoing brain activity may be better modeled by the class of multifractal (MF) processes. This class contains and extends that of self similar processes (e.g.…”

Scale-Free Properties of the Functional Magnetic Resonance Imaging Signal during Rest and Task

“…These signals have been termed scale-free brain activity because a P ∝1/ f β power spectrum is indicative of scale-invariance (He et al, 2010). Secondly, scale-free properties of the fMRI signal vary among gray matter, white matter and cerebrospinal fluid (Bullmore et al, 2004; Ciuciu et al, 2011) and between brain networks (He et al, 2010), arguing strongly against an instrumental-noise origin. Most importantly, recent evidence showing that the power-law exponent of brain field potentials decreases during task-activation (He et al, 2010) suggests that scale-free brain activity is functionally significant.…”

Learning-induced modulation of scale-free properties of brain activity measured with MEG

“…turbulence, stock market...) [1], this infraslow activity has long been attributed to sensor or neural noise and considered to be functionally irrelevant. However, since the last decade, a growing body of evidence has shown a modulation of the 1/f slope between contrasted cognitive states (awake vs phases of sleep [1,2], task vs rest [3,4]) and in pathologies [5] using different imaging techniques (ECoG [1,6], EEG [7][8][9], MEG [8,10] and fMRI [3,6,11,12]). Such results suggest that the infraslow activity also carries meaningful information for brain function.…”