Functional and Developmental Significance of Amplitude Variance Asymmetry in the BOLD Resting-State Signal Ben Davis 1 , Jorge Jovicich 1,2 , Vittorio Iacovella 1 and Uri Hasson 1,2 1 Center for Mind/Brain Sciences (CIMeC), University of Trento, I-38060 Mattarello (TN), Italy and 2 Department of Psychology and Cognitive Sciences, University of Trento, Trento, Italy Address correspondence to Uri Hasson. Email: uri.hasson@unitn.it It is known that the brain’s resting-state activity (RSA) is organized in low frequency oscillations that drive network connectivity. Recent research has also shown that elements of RSA described by high-frequency and nonoscillatory properties are non-random and functionally relevant. Motivated by this research, we investigated nonoscillatory aspects of the blood-oxygen-level-dependent (BOLD) RSA using a novel method for characterizing subtle fluctuation dy- namics. The metric that we develop quantifies the relative variance of the amplitude of local-maxima and local-minima in a BOLD time course (amplitude variance asymmetry; AVA). This metric reveals new properties of RSA activity, without relying on connectivity as a descriptive tool. We applied the AVA analysis to data from 3 differ- ent participant groups (2 adults, 1 child) collected from 3 different centers. The analyses show that AVA patterns a) identify 3 types of RSA profiles in adults’ sensory systems b) differ in topology and pattern of dynamics in adults and children, and c) are stable across magnetic resonance scanners. Furthermore, children with higher IQ demonstrated more adult-like AVA patterns. These findings indicate that AVA reflects important and novel dimensions of brain develop- ment and RSA. Keywords: amplitude asymmetry, development, nonlinear, nonstationary, resting-state Introduction Understanding the organizing features of resting-state brain activity is one of the dominant themes in neuroscientific investigation of human cognition. Since Biswal et al.’s (1995) pioneering work demonstrating correlated spontaneous low- frequency blood-oxygen-level-dependent (BOLD) oscillations between the left and right motor cortices during rest, many studies have identified large, distributed, and functionally related brain regions whose low-frequency BOLD signals similarly oscillate in the absence of an explicit task. Some of the most prominent examples of these resting-state networks are the ventral and dorsal attention networks (Fox et al. 2006), a default-mode network (DMN) including midline and inferior parietal regions (Buckner et al. 2008), networks within auditory and visual cortices (Cordes et al. 2000), and a basal-ganglia network (Robinson et al. 2009). The topology of these networks exhibit “small-world” features (Achard et al. 2006). The size of the networks and the strength of the corre- lations that bind them have been linked to task performance (Hampson et al. 2006), disease (for a review, see Zhang and Raichle 2010), development (Fair et al. 2008), and genetic background (Glahn et al. 2010). More recently, and conjointly with some of these develop- ments, a somewhat divergent line of research has begun to show that resting-state activity (RSA) contains information that is independent of what can be observed by studying low- frequency BOLD oscillations or the networks that they main- tain. Specifically, several studies have shown the variance of BOLD fluctuations within short time frames, which cannot be driven by low-frequency components, contain information important for understanding RSA (e.g., Fair, Schlaggar et al. 2007; Hasson et al. 2009; Garrett et al. 2010) as well as non- random patterns of BOLD spikes during rest (Morgan et al. 2008; Petridou et al. forthcoming; Tagliazucchi et al. 2010, 2012). Developing on this framework, we present a novel ap- proach for studying dynamic components of BOLD RSA. As we show, this method affords new understanding of BOLD dynamics during rest, differentiates between modes of RSA not identified to date and is correlated with both age and IQ. In what follows we briefly review empirical evidence high- lighting the importance of BOLD signal features that extend beyond low-frequency–driven connectivity. We then present the logic of the current method, which is based on studying peak asymmetry in BOLD resting-state signals and its appli- cation to 3 resting-state datasets. After presenting the results, we discuss their implications for understanding RSA and relate them to recent approaches for understanding brain activity in terms of dynamical systems. Information Contained in Higher Frequency Fluctuations Recent work strongly supports the view that there is more information in RSA than low-frequency oscillation. As detailed below, sporadic bursts of strong activity that have typically been considered as noise has been shown to carry meaningful information. Furthermore, studies that have employed pro- cedures where splices from larger sections of time series are concatenated (thus reducing the impact of low-frequency oscillations) have demonstrated that variance within these splices carries meaningful information and is associated with functional connectivity. For example, Garrett et al. (2010) showed that the variance of the BOLD signal within short 20 s fixation epochs correlates with chronological age. Fair, Schlaggar et al. (2007) found that established resting-state connectivity patterns are obtained by concatenating short BOLD time series collected between tasks blocks (30 s, 17 s splices). Similarly, Hasson et al. (2009) found that BOLD activity during short rest epochs (16 s) was synchronized across cortical regions, indicating that this covariance was not driven by low-frequency components. These findings are supported by several studies showing that an important aspect of RSA consists of spontaneous bursts of BOLD activity whose profiles are similar to hemody- namic responses evoked by external stimuli. Patterns of spon- taneous BOLD spikes during rest have been documented in the primary visual cortex (Wang et al. 2008). Spikes were defined as time points where the BOLD magnitude exceeded © The Author 2013. Published by Oxford University Press. All rights reserved. For Permissions, please e-mail: journals.permissions@oup.com Cerebral Cortex doi:10.1093/cercor/bhs416 Cerebral Cortex Advance Access published January 16, 2013 at Universita degli Studi di Trento on February 12, 2013 http://cercor.oxfordjournals.org/ Downloaded from