During rest, the human brain performs essential functions such as memory maintenance, which are associated with resting-statebrainnetworks (RSNs) including the default-mode network (DMN) and frontoparietal network (FPN). Previous studies based on spiking-neuron network models and their reduced models, as well as those based on imaging data, suggest that resting-state network activity can be captured as attractor dynamics, i.e., dynamics of the brainstate toward an attractive state and transitions between different attractors. Here, we analyze the energylandscapes of the RSNs by applying the maximum entropy model, or equivalently the Ising spin model, to human RSN data. We use the previously estimated parameter values to define the energy landscape, and the disconnectivity graph method to estimate the number of local energy minima (equivalent to attractors in attractor dynamics), the basin size, and hierarchical relationships among the different local minima. In both of the DMN and FPN, low-energy local minima tended to have large basins. A majority of the network states belonged to a basin of one of a few local minima. Therefore, a small number of local minima constituted the backbone of each RSN. In the DMN, the energy landscape consisted of two groups of low-energy local minima that are separated by a relatively high energy barrier. Within each group, the activity patterns of the local minima were similar, and different minima were connected by relatively low energy barriers. In the FPN, all dominant local minima were separated by relatively low energy barriers such that they formed a single coarse-grained global minimum. Our results indicate that multistable attractor dynamics may underlie the DMN, but not the FPN, and assist memory maintenance with different memory states. Keywords: resting-state network, maximum entropy model, Ising model, attractor dynamics, functional connectivity
The present findings extend previous results from our group revealing reduced local efficiency of short-range connections among left temporo-parietal sensors in the β3 band (Dimitriadis et al., 2013). Evidence suggesting a poorly integrated sensor-level network among children with dyslexia has also been reported based on minimal spanning tree analyses of resting-state EEG data (Gonzalez et al., 2016). Although not directly comparable, these results are generally consistent with reports of disrupted network structure and various connectivity abnormalities in dyslexia (Frye et al., 2012; Finn et al., 2014; Koyama et al., 2010; 2013). Additional aberrant features of CFC interactions among RD students were highlighted in this study, representing globally reduced long-range CFC interactions compared to non-impaired readers. It has been proposed that cross-frequency interactions support the synchronization of nested hierarchical activities of neuronal assemblies oscillating on a dominant frequency mode (Buzsáki, 2010). This mechanism purportedly supports the accuracy in the timing of exchanged information among different oscillatory rhythms and the dynamic control of anatomically distributed functional networks (Buzsáki, 2006; Canolty and Knight, 2010).
In the present study, it is demonstrated that there were increased and decreased FCS of brainnetworks in DR patients. Decreased FCS (patients group compared with controls group) involved the nodes of visual cortex, audi- tory cortex, and limbic system, while increased FCS involved the nodes mainly in the frontal lobe and cingu- lum, which forms a circuit containing disconnection and compensatory networks. It emphasizes on the underlying circuit differing from the former study on the correlation study between brainnetworks and clinical indexes. 4 There are some new insights into retinal physiology, suggesting that the retinal dysfunction in DR patients may be viewed as a change in the retinal neurovascular unit. 15 The neuro- vascular unit is supposed to be altered in the patients with DR, with changes in neural function and neurotransmitter metabolism and loss of blood – brain barrier. The neural function changes may be the critical underlying mechan- ism of brain network alterations in DR patients and it may elucidate the physical basis of the brain network circuit.
While decades of neuroscientific research has detailed the brainnetworks underlying memory, to date the neurobiology underlying interindividual memory differences in a healthy population is not known. Here we use the behavioral and restingstate fMRI data from the Human Connectome Project (HCP), and predict subjects’ scores on tests of working and episodic memory based on their whole brain functional connectivity significantly above chance. We observed that brain connectivity between regions determining differences between healthy subjects were different from those traditionally associated with memory. Results may ultimately be relevant to determine risk factors for the development of neurodegenerative disorders.
Graph methods provide a distinct alternative to seed-based and ICA methods [48,60 – 63]. This approach views RSNs as a collection of nodes connected by edges (Fig. 2, 3rd panel). Here, the relation between the nodes and edges can be established as G ¼ ð V : E Þ where V is a gathering of nodes connected by edges E, which describes the interaction between nodes. In this approach, ROIs are represented by nodes and the correlations among the ROIs are demonstrated as the level of connectivity (weights) using the edges. The characteristics of the graph can be evaluated to quantify the distribution of functional hubs (highly functionally connected nodes) in the human brain [38,63]. Examples of measures of interest include: (i) average path length; (ii) clustering coef ﬁ cient; (iii) nodal degree; (iv) centrality measures; and (v) level of modularity . Using the graph theory technique, several studies have demonstrated that the brain exhibits a small world topology. Small world topology was ﬁ rst described in social networks. It allows each node to have a relatively low number of connections while still being connected to all other nodes within a short distance (that is, short distances between any two nodes). The small world is achieved through the existence of hubs, which are critical nodes with large number of connections, allowing a high level of local connectivity (neighboring nodes) . While seed- based analysis focuses only on the strength of correlation between one ROI to another, graph theory measures the topological properties of an ROI within the whole brain or the network related to a particular function. It has exhibited good correspon- dence with well-known anatomical white matter tracts (structural highways of the brain) and resting-statenetworks .
Our results provide new insight into the rich dynamic organization of restingstate in terms of network components that segregate and integrate over short time intervals. The iCAPs form a compact set of building blocks that can be ﬂexibly combined to describe spontaneous activity ﬂuctuations. The combinations of iCAPs observed are not arbitrary, and we only observed 4% of all possible combinations. Clustering of iCAPs on the basis of these combinations shows a hierarchical organization of large-scale brainnetworks that is consistent with differences in behaviour proﬁles of iCAPs; that is, resting-state activity is unravelled into periods when components related to sensory, default-mode and attention, respectively, are dominating (Fig. 6). The DMN is the hallmark of the brain’s restingstate and has been referred to as the main hub of the internal mode of cognition, related to higher-level processes such as memory and learning 37,38 . Our analysis enables us to study both spatial and temporal interactions of consistent co-activation patterns, including the DMN. The DMN iCAP (8) has the strongest spatial similarity with the conventional DMN as determined by PCC-seed-based connectivity, followed by precuneus (5), pDMN (10) and ACC (13), which represent subnetworks of the DMN; the negative part of the DMN PCC-seed connectivity map correlates with attention (2) and anterior salience (11, 14). Spatial subdivisions can also be obtained for other seed-based connectivity maps, such as primary visual and motor networks (see Supplementary Fig. 12).
Photoacoustic imaging of the brain is based on the acoustic detection of optical absorption from tissue chromophores, such as oxy-hemoglobin (HbO) and deoxy-hemoglobin (HbR) [190,191]. PAI can simultaneously provide high-resolution images of the brain vasculature and hemodynamics [192,193]. Photoacoustic technique is a scalable imaging technique, i.e., from photoacoustic microscopy (PAM) to photoacoustic tomography (PAT). PAM is an appealing imaging modality for shallow targets as it gives detailed information of vascular, functional, and molecular aspects [194,195] of biological samples both ex-vivo and in-vivo . PAT is used for deep tissue imaging applications where coarse resolution ( 50 to 200 m m) is acceptable. PAT can potentially be used for neonatal brain imaging. In PAT, a diffused high energy pulsed laser, through bulk optics or optical ﬁ ber bundle, covers the imaging target, e.g., head, and photoacoustic waves are generated. The waves around the tissue are collected by wideband ultrasound transducers (Fig. 9) . The detection scheme can be realized by a linear array ultrasound transducer, a single ultrasound transducer which rotates around the sample, by spiral rotation of a transducer, by a stationary ring array of 128, 256, or by a number of transducers arranged as an arc array transducer or as a half or full hemispherical array. For neonatal brain imaging, a linear or a hemispherical transducer array can be used. The ultrasound waves collected from the object are measured and sent to a computer to be used by an image reconstruction algorithm , to form an image; in the case of hemispherical transducer array, a three dimensional map of brain activities is obtained .
The most relevant group- level IC maps (of 34) were selected according to the following three steps. First, group- level IC maps with more than 33% of the estimated spectral power in high frequencies (>0.1 Hz) were excluded to keep only networks within the low- frequency range of 0.1–0.01 Hz (Lowe, Mock, & Sorenson, 1998; Tyszka et al., 2014). Second, Smith et al. (2009) described the major covarying networks in the restingbrain and created a template of these RSNs widely used in resting- state fMRI studies. With this template and our remaining group maps, a function, using the “goodness- of- fit” approach was created and applied (Greicius, Srivastava, Reiss, & Menon, 2004; Vanhaudenhuyse et al., 2010). Finally, the third step consisted in a visual inspection of each component spatial profile to verify the con- sistency and ensure the effectiveness of the two previous steps. Plus, this last step allows us to select other known and well- described net- works that are not in Smith and colleagues’ template, but still comply with first selection step.
For any given task, a host of distributed, functionally specialized brain areas work in concert to integrate sensorial inputs with previously stored information, as well as with executive and motor regions to generate an appropriate behavior. The set of brain regions that interact in this manner make up large-scale functional networks . A network perspective of brain function, accounting for the interactions between regions, offers a potentially useful framework for the study of normal functioning, and also for the identification of relevant intermediate pathological phenotypes . Despite being in its early stages, the network approach applied to PD has shown potential clinical usefulness as a tool for differential diagnosis, monitoring disease progression and treatment response, and also for the development of biomarkers for complications such as dementia. Non-invasive in vivo neuroimaging techniques also offer an unprecedented opportunity to characterize the pathophysiological substrates underlying different manifestations of the disease.
Synchronization likelihood was computed on the pruned EEG datasets for all pair-wise combination of the channels, yielding a 14 × 14 weighted synchronization matrix for every time point, in which the connection strength is assumed to be proportional to the level of synchronization between brain regions. Each of these matrices capture the actual topology of the underlying network, and calculating different network measures over them yield Network Metrics Time Series (NMTS) that describe the temporal evolution of network topology. Complex networks have several aspects to their topologies such as modularity or small-worldness ( Rubinov and Sporns, 2010 ). The network is required to contain a sufficiently large number of nodes for network descriptors to make sense, i.e., there is no point in calculating for example the node degree distribution on a network with 14 nodes. It has been demonstrated however, that global network measures Density (D), Clustering Coefficient (C) and Efficiency (E) can be used effectively to describe and capture significant topological differences in smaller networks ( Racz et al., 2017 ). We used the weighted formulas to calculate the aforementioned network measures. Weighted Density (often termed also Connectivity Strength) is the fraction of overall connectivity strength present to the maximal possible connection strength in a network ( Rubinov and Sporns, 2010 ) and calculates as
When investigating functional connectivity patterns of the brain, researchers often use so-called resting-state protocols. Instead of looking at task-evoked activations, resting-state data is purported to reveal the intrinsic activation patterns of the brain while the participants are scanned without performing any tasks, i.e. at “rest” (Biswal, Yetkin, Haughton, & Hyde, 1995). It is due to this rest condition that it is believed that this method elucidates the funda- mental organisation of the brain, as neural activity is allowed to fluctuate unconstrained (Cole et al., 2010). Several studies have documented reliable brainnetworks during rest across vari- ous psychological states (Greicius et al., 2008; Horovitz et al., 2009; Horovitz et al., 2008), across species (Pawela et al., 2008; Rilling et al., 2007; Vincent et al., 2007), and across human subjects (Friston, 2009; Gusnard & Raichle, 2001; Shehzad et al., 2009). This strongly indicates that the functional hierarchical organisation of the brain is an evolutionary conserved trait. Furthermore, it has been shown that the correlation patterns of slow oscillations (< 0.1 Hz) spontaneously occurring at rest form widely distributed resting-state functional networks (RSNs), such as the sensory systems (De Luca, Beckmann, De Stefano, Matthews, & Smith, 2006), the motor cortex (Biswal et al., 1995), the attention systems (Fox, Corbetta, Snyder, Vincent, & Raichle, 2006), and the memory systems (Greicius, Krasnow, Reiss, & Menon, 2003). In a review of resting-state literature Fox & Raichle (2007) argue that inherent BOLD changes correlate with variability in human behaviour, and the authors propose that the in- trinsic co-activation plays an important role in cognition and behaviour, perhaps through this fundamental functional organisation of the brain.
At first glance, subsystems with high functional integration are also expected to display high functional segregation. The fact that Connectome subsystems have relatively high values of both Subsystem Integration and Subsystem-Environment MI suggests that they may balance a trade-off between two important information-processing functions: accessing information from large areas of the brain and integrating it efficiently across a network of hub regions ( Zamora-López et al., 2010 ). We investi- gated this question by looking at particular values of Subsystem Integration and Subsystem-Environment MI for subsystems of size of 11 ( ∼5% of one hemisphere) (Figure 7A). We chose Connectome subsystems with high values on both Subsystem- Integration and Subsystem-Environment MI and found that they are distributed into four minimally-overlapping subsystem com- munities (Figure 7B). Interestingly, these communities can be interpreted in terms of neural anatomy as well as in terms of previous work on functional restingstatenetworks. The yellow communities in the left and right hemispheres occupy areas corre- sponding to primary and secondary visual and auditory cortices, the light blue communities roughly correspond to locations in the default mode network, while the dark red and blue communities contain regions reported to be part of the ventral attention, dor- sal attention, and fronto-parietal control restingstatenetworks ( Yeo et al., 2011 ). The anatomical regions (Figure 7C) most rep- resented in the light blue, dark blue and dark red communities are known to include many functional hub regions, such as supe- rior frontal gyrus, inferior and superior parietal lobules, supra- marginal gyrus and insula. Interestingly, in both hemispheres, the superior frontal gyrus included ROIs corresponding to all three of these communities, suggesting that it may be a location where these separate high-level integrative networks intersect.
Early brain development is characterized by rapid growth and perpetual reconfiguration, driven by a dynamic milieu of heterogeneous processes. Postnatal brain plasticity is associated with increased vulnerability to environmental stimuli. However, little is known regarding the ontogeny and temporal manifestations of inter- and intra-regional functional connectivity that comprise functional brainnetworks. Resting-state functional magnetic resonance imaging (rs-fMRI) has emerged as a promising non-invasive neuroinvestigative tool, measuring spontaneous fluctuations in blood oxygen level dependent (BOLD) signal at rest that reflect baseline neuronal activity. Over the past decade, its application has expanded to infant populations providing unprecedented insight into functional organi- zation of the developing brain, as well as early biomarkers of abnormal states. However, many methodological issues of rs-fMRI analysis need to be resolved prior to standard- ization of the technique to infant populations. As a primary goal, this methodological manuscript will (1) present a robust methodological protocol to extract and assess rest- ing-statenetworks in early infancy using independent component analysis (ICA), such that investigators without previous knowledge in the field can implement the analysis and reliably obtain viable results consistent with previous literature; (2) review the current methodological challenges and ethical considerations associated with emerging field of infant rs-fMRI analysis; and (3) discuss the significance of rs-fMRI application in infants for future investigations of neurodevelopment in the context of early life stressors and pathological processes. The overarching goal is to catalyze efforts toward development of robust, infant-specific acquisition, and preprocessing pipelines, as well as promote greater transparency by researchers regarding methods used.
Changes in whole-brain connectivity networks between task and rest
In the previous section we addressed how connections within the DMN network change with rest state and task. Here, we are interested in determining how large-scale networks spanning the entire cortex are altered in the task-condition compared to the restingstate-condition. Identifying properties of the “whole brain” network allows us to examine the overall structure of neuronal communication among spatially precise regions. To study connections spanning the cortex, we defined 90 anatom- ical ROIs based on Brodmann Areas spanning both the right and left cortical hemispheres using the WFU_PickAtlas toolbox . Connections among these ROIs were used to compute network statistics to capture global tendencies of connectivity in resting- state and task networks, and to characterize the consistencies and differences in the behavior of individual cortical areas us- ing graph theoretical measures. Figure 4 illustrates the whole- brainnetworks of the restingstate (Figure 4A), and task (Figure 4B) networks with Brodmann areas as the ROIs. The connections represent bivariate correlations with p<0.05 (FDR Corrected), and node color indicates the difference between the positive and negative connections (green to dark blue indicating increasing negative total connections respectively, green to dark red indi- cating increasing positive total connections). In the following sec- tions, we discuss the quantitative properties of these networks. Degree distribution
Resting-state fMRI (rs-fMRI) has been widely used to explore the functional coupling between distinct brain regions by calculating low-frequency spontaneous fluc- tuations in the time series, i.e., functional connectivity (Biswal et al. 1995; Fox and Raichle 2007; Buckner et al. 2013; Song and Jiang 2012). Functional connec- tivity has been a powerful tool to identify the resting- statenetworks (Greicius et al. 2003; Tomasi and Volkow 2012; Damoiseaux et al. 2006). Recently, rs-fMRI has also been exploited to delineate distinct subregions within a larger brain region based on differential patterns of functional connectivity (Craddock et al. 2012; Kim et al. 2010; Nelson et al. 2010; Yeo et al. 2011; Shen et al. 2010; Deen et al. 2011). As we know, the func- tional connectivity could be influenced by various arti- facts in rs-fMRI data including physiological artifacts (Birn et al. 2008), transient head motion (Van Dijk et al. 2012), different scanning conditions (Patriat et al. 2013) and preprocessing procedures (Van Dijk et al. 2010; Satterthwaite et al. 2013). Hereby, these artifacts might also have impacts on the parcellation results. Generally, there are three approaches proposed to reduce the impact of noise during the parcellation procedures. The first is to average the connectivity profiles (Deen et al. 2011; Yeo et al. 2011) or similarity matrices across subjects (Craddock et al. 2012), which, however, eliminates inter- individual variability, which has been widely reported in both structure and function of the human brain (Mueller et al. 2013; Rademacher et al. 2001; Zilles and Amunts 2013). The second is to employ spatial constraints to improve the stability of parcellation (Craddock et al. 2012), which might bias the results towards spherical- shaped clusters. Another approach is to remove the noisy edges lying between clusters by constructing a sparse similarity matrix, for instance the KNN graph (Shen et al. 2010; von Luxburg 2007). But the KNN graph method requires a global sparsity parameter, which is often difficult to determinate (Nadler and Galun 2006) and could significantly affect the performance of par- cellation (Shen et al. 2010). Thus, a more efficient sparse technique is required, which could generate robust brain parcellation by guaranteeing the stability of parcellation and retaining the individual variability at the same time. The sparse representation theory (Elad 2010) has been widely employed in the classification of face, natural and medical images (Wright et al. 2009, 2010; Su et al. 2012; Wee et al. 2014; Mairal et al. 2008). Recently, it also has been proposed for data clustering and achieved robustness on high-dimensional data (Elhamifar and Vi- dal 2013), which construct a sparse similarity graph based on the sparse representation coefficients and
The overall results are therefore consistent with a pattern of spatially more restricted local connectivity in mature adult RSNs, possibly indicative of more functional specialization with greater maturity. It was not a ubiquitous observation, however, as, for example, within the frontoparietal executive control networks, there were no areas in which infant group connectivity exceeded that of the adult group. In contrast, adults had higher connectivity evident in both left and right frontal and posterior components of the executive networks. Also, within the language network, infants were observed to have greater connectivity within the left STG, certainly a region classically associated with language function, whereas adults had no language network areas of significantly greater connectivity.
To our knowledge, this is the first functional connectivity study devoted to the RN and, incidentally, to the SN in hu- mans. During the brainrestingstate, the RN displays strong functional coherence with associative prefrontal, insular, tem- poral, and parietal cortices; the thalamus; and the hypothala- mus, but not with the sensorimotor cortex. These results sup- port a cognitive role of the RN, probably related to salience detection and executive control. This rubral circuit, which seems to constitute a modular cerebellar subsystem, clearly differs from the striatal loops passing through the very close and anatomically associated SN. Our study also confirms pre- vious results for the SN obtained in monkeys, reporting corti- conigral connections with the prefrontal, temporal, and occip- ital cortices. Finally, our study demonstrates that this fcMRI technique can be successfully applied to small subcortical structures. However, further investigation is warranted to fully elucidate the specific clinical significance of the rubral and nigral networks that have been demonstrated in this pre- liminary and exploratory study.
reactivity to distinct stimuli, as well as age- and/ or cognitive-level-dependent response (Graham, Pfeifer et al. 2015). Such variables often impede extrapolation of findings to other age groups, critical for investigations of longitudinal neurodevelopment. The emergence of infant resting-state fMRI in the past decade offers unmitigated insight into patterns of functional connectivity, yielding more complete representations of neural networks and their development. First described in Fransson’s seminal paper (Fransson, Skiold et al. 2007) , the presence of RSNs in infants have been established as early as the fetal (Thomason, Grove et al. 2015, Thomason, Scheinost et al. 2017), preterm (Fransson, Skiold et al. 2007, Doria, Beckmann et al. 2010, He and Parikh 2016) and infant periods (Fransson, Skiold et al. 2009, Gao, Gilmore et al. 2013, Wylie, Rojas et al. 2014), undergoing substantial maturation and refinement over the first two decades of life (Fair, Dosenbach et al. 2007, Fair, Cohen et al. 2008, Gao, Zhu et al. 2009, Smyser, Inder et al. 2010, Thomason, Dennis et al. 2011). Despite limited published literature since its inception in 2007, significant groundwork has been laid in the field of infant resting-state fMRI, offering transient glimpses into the complex interplay of structural and functional brain development.
Finally, we analyzed the variability of between-network information lows. To that end, we computed the log ratios of the variability between the aferent and eferent lows in each pair of networks. Speciically, for each two networks A and B, we calculated the natural logs of the average variance from A to B and that from B to A (see Fig. 5 (e) ). he derived log values were further divided by each other to generate a ratio. Next, we conducted a pair-wise ANOVA test to compare the log ratios of every pair of networks with that of the whole brain to exam- ine whether the diferences between variability of information inlow and outlow were statistically signiicant. Our results showed that, compared to the variability of information low in the whole brain, there was signii- cantly higher information low variability among medial frontal cortex, frontal parietal regions, the default mode regions (all P FWE ’s between each two networks < 0.001), suggesting that information low between association
Light is considered to modulate human brain function only via the retinal pathway, a way of thinking that we aimed to challenge in the present study. Literature provides evidence of inherent phototransduction for instance in the rat brain and there are potentially photosensitive opsin proteins like melanopsin and panopsin in the human brain too. In order to investi- gate a short term response, functional connectivity changes of the brain were studied in the restingstate with functional magnetic resonance imaging during bright light stimulus via the ear canal. Lateral visual and sensorimotor networks showed increased func- tional connectivity in the light stimulus group com- pared to sham controls. The lateral visual network demonstrated slowly increasing functional connec- tivity on average and the same temporal characteris- tic was shared by diverse cerebellar brain regions. Hypothetical phototransduction signal pathways lead- ing to responses in brain function are discussed as well as some observed effects and their possible link to the findings. Findings from this study together with the plausible photoreceptor candidates suggest that the brain possesses photosensitive properties, which will have interesting implications for the modulation of brain function and understanding the basic physiol- ogy of the brain.