The relation between large-scale brain structure and function is an outstanding open problem in neuroscience. We approach this problem by studying the dynamical regime under which realistic spatio-temporal patterns of brain activity emerge from the empirically derived network of human brain neuroanatomical connections. The results show that critical dynamics unfolding on the structural connectivity of the human brain allow the recovery of many key experimental findings obtained with functional Magnetic Resonance Imaging (fMRI), such as divergence of the correlation length, anomalous scaling of correlation fluctuations, and the emergence of large-scale resting state networks.Understanding the relation between brain architecture and function is a central question in neuroscience. In that direction, important efforts over recent years have been devoted to map the large-scale structure of the human cortex, including attempts to build brain structural connectivity matrices from imaging data. An example is the connectivity matrix of the entire human brain, recently derived from fiber densities measured between a large number (500-4000) of homogeneously distributed brain regions [1]. This and related work encompasses a large collaborative project dubbed the brain "connectome" [3], whose ultimate goal is to understand in detail the architecture of whole-brain connectivity. However, "like genes, structural connections alone are powerless", thus "the connectome must be expressed in dynamic neural activity to be effective in behavior and cognition" [2]. The results presented in this Letter show that very relevant aspects of brain dynamics can be predicted from the structure provided that the underlying dynamics are critical.To guide our comparison with available experimental results, we choose to concentrate on robust findings concerning brain dynamics. Specifically, we ask how spontaneous brain dynamics at the large scale organize into the relatively few spatio-temporal patterns revealed experimentally in recent years [4]. This is important because a wide range of experiments using functional Magnetic Resonance Imaging (fMRI) have emphasized that these spatial clusters of coherent activity, termed Resting State Networks (RSN) [5], are specifically associated with neuronal systems responsible for sensory, cognitive and behavioural functions [6]. Furthermore, the pattern of correlations in these networks has been shown to change with various cognitive and pathophysiological conditions [4]. Of interest here are studies showing that the RSN activity exhibits peculiar scaling properties, resembling dynamics near the critical point of a second order phase transition [7][8][9], consistent with evidence showing that the brain at rest is near a critical point [10]. These empirical findings are in line with computational modeling results [11][12][13].Here we study whether a simple dynamical model running over the empirical structure of neuroanatomical connections [1] suffices to replicate the aforementioned fundamental features of ...

Abstract. We discuss a Bayesian model selection approach to high dimensional data in the deep under sampling regime. The data is based on a representation of the possible discrete states s, as defined by the observer, and it consists of M observations of the state. This approach shows that, for a given sample size M , not all states observed in the sample can be distinguished. Rather, only a partition of the sampled states s can be resolved. Such partition defines an emergent classification q s of the states that becomes finer and finer as the sample size increases, through a process of symmetry breaking between states. This allows us to distinguish between the resolution of a given representation of the observer defined states s, which is given by the entropy of s, and its relevance which is defined by the entropy of the partition q s . Relevance has a non-monotonic dependence on resolution, for a given sample size. In addition, we characterise most relevant samples and we show that they exhibit power law frequency distributions, generally taken as signatures of "criticality". This suggests that "criticality" reflects the relevance of a given representation of the states of a complex system, and does not necessarily require a specific mechanism of self-organisation to a critical point.

Human brain dynamics and functional connectivity fluctuate over a range of temporal scales in coordination with internal states and environmental demands. However, the neurobiological significance and consequences of functional connectivity dynamics during rest have not yet been established. We show that the coarse-grained clustering of whole-brain dynamic connectivity measured with magnetic resonance imaging reveals discrete patterns (dynamic connectivity states) associated with wakefulness and sleep. We validate this using EEG in healthy subjects and patients with narcolepsy and by matching our results with previous findings in a large collaborative database. We also show that drowsiness may account for previous reports of metastable connectivity states associated with different levels of functional integration. This implies that future studies of transient functional connectivity must independently monitor wakefulness. We conclude that a possible neurobiological significance of dynamic connectivity states, computed at a sufficiently coarse temporal scale, is that of fluctuations in wakefulness.

Synchronization of brain activity fluctuations is believed to represent communication between spatially distant neural processes. These inter-areal functional interactions develop in the background of a complex network of axonal connections linking cortical and sub-cortical neurons, termed the human "structural connectome".Theoretical considerations and experimental evidence support the view that the human brain can be modeled as a system operating at a critical point between ordered (sub-critical) and disordered (super-critical) phases. Here, we explore the hypothesis that pathologies resulting from brain injury of different etiology are related to the model of a critical brain. For this purpose, we investigate how damage to the integrity of the structural connectome impacts on the signatures of critical dynamics. Adopting a hybrid modeling approach combining an empirical weighted network of human structural connections with a conceptual model of critical dynamics, we show that lesions located at highly transited connections progressively displace the model towards the subcritical regime. The topological properties of the nodes and links are of less importance when considered independently of their weight in the network. We observe that damage to midline hubs such as the middle and posterior cingulate cortex is most crucial for the disruption of criticality in the model. However, a similar effect can be achieved by targeting less transited nodes and links whose connection weights add up to an equivalent amount. This implies that brain pathology does not necessarily arise due to insult targeted at well-connected areas and that inter-subject variability could obscure lesions located at non-hub regions. Finally, we discuss the predictions of our model in the context of clinical studies of traumatic brain injury and neurodegenerative disorders.3

Risky decision-making seems to be markedly disrupted in patients with chronic pain, probably due to the high cost that impose pain and negative mood on executive control functions. Patients’ behavioral performance on decision-making tasks such as the Iowa Gambling Task (IGT) is characterized by selecting cards more frequently from disadvantageous than from advantageous decks, and by switching often between competing responses in comparison with healthy controls (HCs). In the present study, we developed a simple heuristic model to simulate individuals’ choice behavior by varying the level of decision randomness and the importance given to gains and losses. The findings revealed that the model was able to differentiate the behavioral performance of patients with chronic pain and HCs at the group, as well as at the individual level. The best fit of the model in patients with chronic pain was yielded when decisions were not based on previous choices and when gains were considered more relevant than losses. By contrast, the best account of the available data in HCs was obtained when decisions were based on previous experiences and losses loomed larger than gains. In conclusion, our model seems to provide useful information to measure each individual participant extensively, and to deal with the data on a participant-by-participant basis.

We propose an end to end approach using graph construction and semantic representation learning to solve the problem of structured information extraction from heterogeneous, semi-structured, and high noise human readable documents. Our system first converts PDF documents into single connected graphs where we represent each token on the page as a node, with vertices consisting of the inverse euclidean distances between tokens. Token, lines, and individual character nodes are augmented with dense text model vectors. We then proceed to represent each node as a vector using a tailored GraphSAGE algorithm that is then used downstream by a simple feedforward network. Using our approach, we achieve state-of-the-art methods when benchmarked against our dataset of 205 PDF invoices. Along with generally published metrics, we introduce a highly punitive yet application specific informative metric that we use to further measure the performance of our model.