Connectivity Matrices Research Articles

Epilepsy-Related Brain Network Alterations in Patients With Temporal Lobe Glioma in the Left Hemisphere.

Background: Seizures are a common symptom in patients with temporal lobe gliomas and may result in brain network alterations. However, brain network changes caused by glioma-related epilepsy (GRE) remain poorly understood.Objective: In this study, we applied graph theory analysis to delineate topological networks with resting-state functional magnetic resonance images (rs-fMRI) and investigated characteristics of functional networks in patients with GRE.Methods: Thirty patients with low-grade gliomas in the left temporal lobe were enrolled and classified into GRE (n = 15) and non-GRE groups. Twenty healthy participants matched for age, sex, and education level were enrolled. All participants had rs-fMRI data. Sensorimotor, visual, default mode, auditory, and right executive control networks were used to construct connection matrices. Topological properties of those sub-networks were investigated.Results: Compared to that in the GRE group, four edges with higher functional connectivity were noted in the non-GRE group. Moreover, 21 edges with higher functional connectivity were identified in the non-GRE group compared to the healthy group. All significant alterations in functional edges belong to the visual network. Increased global efficiency and decreased shortest path lengths were noted in the non-GRE group compared to the GRE and healthy groups. Compared with that in the healthy group, nodal efficiency of three nodes was higher in the GRE and non-GRE groups and the degree centrality of six nodes was altered in the non-GRE group.Conclusion: Temporal lobe gliomas in the left hemisphere and GRE altered visual networks in an opposing manner. These findings provide a novel insight into brain network alterations induced by GRE.

Eigenvalues of Ising connection matrix with long-range interaction

We examine multidimensional Ising systems on hypercube lattices and calculate analytically the eigenvalues of their connection matrices. We express the eigenvalues in terms of spin–spin​ interaction constants and the eigenvalues of the one-dimensional Ising connection matrix (the latter are well known). To do this we present the eigenvectors as Kronecker products of the eigenvectors of the one-dimensional Ising connection matrix. For periodic boundary conditions, it is possible to obtain exact results for interactions with an arbitrary large number of neighboring spins. We present exact expressions for the eigenvalues for two- and three-dimensional Ising connection matrices accounting for the first five coordination spheres (that is interactions up to next-next-next-next nearest neighbors). In the case of free-boundary systems, we show that in the two and three dimensions the exact expressions could be obtained only if we account for interactions with spins of not more than first three coordination spheres.

Weighted Brain Network Metrics for Decoding Action Intention Understanding Based on EEG.

Background: Understanding the action intentions of others is important for social and human-robot interactions. Recently, many state-of-the-art approaches have been proposed for decoding action intention understanding. Although these methods have some advantages, it is still necessary to design other tools that can more efficiently classify the action intention understanding signals.New Method: Based on EEG, we first applied phase lag index (PLI) and weighted phase lag index (WPLI) to construct functional connectivity matrices in five frequency bands and 63 micro-time windows, then calculated nine graph metrics from these matrices and subsequently used the network metrics as features to classify different brain signals related to action intention understanding.Results: Compared with the single methods (PLI or WPLI), the combination method (PLI+WPLI) demonstrates some overwhelming victories. Most of the average classification accuracies exceed 70%, and some of them approach 80%. In statistical tests of brain network, many significantly different edges appear in the frontal, occipital, parietal, and temporal regions.Conclusions: Weighted brain networks can effectively retain data information. The integrated method proposed in this study is extremely effective for investigating action intention understanding. Both the mirror neuron and mentalizing systems participate as collaborators in the process of action intention understanding.

Alterations in Network Connectivity after Traumatic Brain Injury in Mice.

Victims of mild traumatic brain injury (mTBI) usually do not display clear morphological brain defects, but frequently have long-lasting cognitive deficits, emotional difficulties, and behavioral disturbances. In the present study we used diffusion magnetic resonance imaging (dMRI) combined with graph theory measurements to investigate the effects of mTBI on brain network connectivity. We employed a non-invasive closed-head weight-drop mouse model to produce mTBI. Mice were scanned at two time points, 24 h before the injury and either 7 or 30 days following the injury. Connectivity matrices were computed for each animal at each time point, and these were subsequently used to extract graph theory measures reflecting network integration and segregation, on both the global (i.e., whole brain) and local (i.e., single regions) levels. We found that cluster coefficient, reflecting network segregation, decreased 7 days post-injury and then returned to baseline level 30 days following the injury. Global efficiency, reflecting network integration, demonstrated opposite patterns in the left and right hemispheres, with an increase of right hemisphere efficiency at 7 days and then a decrease in efficiency following 30 days, and vice versa in the left hemisphere. These findings suggest a possible compensation mechanism acting to moderate the influence of mTBI on the global network. Moreover, these results highlight the importance of tracking the dynamic changes in mTBI over time, and the potential of structural connectivity as a promising approach for studying network integrity and pathology progression in mTBI.

Overview on Keloid Disorder: Phenotypic Spectrum, Connective Tissue Pathology, and Treatment Development

Overview on Keloid Disorder: Phenotypic Spectrum, Connective Tissue Pathology, and Treatment Development

Network Autocorrelation Modeling: Bayesian Techniques for Estimating and Testing Multiple Network Autocorrelations

The network autocorrelation model has been the workhorse for estimating and testing the strength of theories of social influence in a network. In many network studies, different types of social influence are present simultaneously and can be modeled using various connectivity matrices. Often, researchers have expectations about the order of strength of these different influence mechanisms. However, currently available methods cannot be applied to test a specific order of social influence in a network. In this article, the authors first present flexible Bayesian techniques for estimating network autocorrelation models with multiple network autocorrelation parameters. Second, they develop new Bayes factors that allow researchers to test hypotheses with order constraints on the network autocorrelation parameters in a direct manner. Concomitantly, the authors give efficient algorithms for sampling from the posterior distributions and for computing the Bayes factors. Simulation results suggest that frequentist properties of Bayesian estimators on the basis of noninformative priors for the network autocorrelation parameters are overall slightly superior to those based on maximum likelihood estimation. Furthermore, when testing statistical hypotheses, the Bayes factors show consistent behavior with evidence for a true data-generating hypothesis increasing with the sample size. Finally, the authors illustrate their methods using a data set from economic growth theory.

A Tale of Two Proteolytic Machines: Matrix Metalloproteinases and the Ubiquitin-Proteasome System in Pulmonary Fibrosis.

Pulmonary fibrosis is a chronic and progressive lung disease characterized by the activation of fibroblasts and the irreversible deposition of connective tissue matrices that leads to altered pulmonary architecture and physiology. Multiple factors have been implicated in the pathogenesis of lung fibrosis, including genetic and environmental factors that cause abnormal activation of alveolar epithelial cells, leading to the development of complex profibrotic cascade activation and extracellular matrix (ECM) deposition. One class of proteinases that is thought to be important in the regulation of the ECM are the matrix metalloproteinases (MMPs). MMPs can be up- and down- regulated in idiopathic pulmonary fibrosis (IPF) lungs and their role depends upon their location and function. Furthermore, alterations in the ubiquitin-proteosome system (UPS), a major intracellular protein degradation complex, have been described in aging and IPF lungs. UPS alterations could potentially lead to the abnormal accumulation and deposition of ECM. A better understanding of the specific roles MMPs and UPS play in the pathophysiology of pulmonary fibrosis could potentially drive to the development of novel biomarkers that can be as diagnostic and therapeutic targets. In this review, we describe how MMPs and UPS alter ECM composition in IPF lungs and mouse models of pulmonary fibrosis, thereby influencing the alveolar epithelial and mesenchymal cell behavior. Finally, we discuss recent findings that associate MMPs and UPS interplay with the development of pulmonary fibrosis.

EEG Connectivity Analysis Using Denoising Autoencoders for the Detection of Dyslexia.

The Temporal Sampling Framework (TSF) theorizes that the characteristic phonological difficulties of dyslexia are caused by an atypical oscillatory sampling at one or more temporal rates. The LEEDUCA study conducted a series of Electroencephalography (EEG) experiments on children listening to amplitude modulated (AM) noise with slow-rythmic prosodic (0.5-1[Formula: see text]Hz), syllabic (4-8[Formula: see text]Hz) or the phoneme (12-40[Formula: see text]Hz) rates, aimed at detecting differences in perception of oscillatory sampling that could be associated with dyslexia. The purpose of this work is to check whether these differences exist and how they are related to children's performance in different language and cognitive tasks commonly used to detect dyslexia. To this purpose, temporal and spectral inter-channel EEG connectivity was estimated, and a denoising autoencoder (DAE) was trained to learn a low-dimensional representation of the connectivity matrices. This representation was studied via correlation and classification analysis, which revealed ability in detecting dyslexic subjects with an accuracy higher than 0.8, and balanced accuracy around 0.7. Some features of the DAE representation were significantly correlated ([Formula: see text]) with children's performance in language and cognitive tasks of the phonological hypothesis category such as phonological awareness and rapid symbolic naming, as well as reading efficiency and reading comprehension. Finally, a deeper analysis of the adjacency matrix revealed a reduced bilateral connection between electrodes of the temporal lobe (roughly the primary auditory cortex) in DD subjects, as well as an increased connectivity of the F7 electrode, placed roughly on Broca's area. These results pave the way for a complementary assessment of dyslexia using more objective methodologies such as EEG.

0143 Vigilance Declines Following Sleep Deprivation are Associated with Two Previously Identified Dynamic Connectivity States

Abstract Introduction Dynamic functional connectivity (DFC) analysis of resting-state fMRI data has been successfully used to track fluctuations in arousal in the human brain. Changes in DFC have also been reported with acute sleep deprivation. Here, we demonstrate that dynamic connectivity states (DCS) previously related to arousal are reproducible, and are associated with individual differences in sustained attention declines after one night of total sleep deprivation. Methods 32 participants underwent two counterbalanced resting-state fMRI scans: during rested wakefulness (RW) and following total sleep deprivation (SD). They also completed the Psychomotor Vigilance Test (PVT), a sustained attention task that is highly sensitive to the effects of sleep loss. SD vulnerability was computed as the decrease in response speed (∆RS) and increase in lapses (∆lapse) in SD compared with RW. Dynamic functional connectivity analysis was conducted on rs-fMRI data. Connectivity matrices were clustered to obtain 5 prototypical DCS. We calculated the proportion of time participants spent in each of these DCS, as well as how often participants transitioned between DCSs. Relationships between SD vulnerability and connectivity metrics were then correlated. Results We recovered two DCS that were highly similar (ρ = .89-.91) to arousal-related DCS observed in previous work (high arousal state (HAS); low arousal state (LAS)). After sleep deprivation, the proportion of time spent in the LAS increased significantly (t29=3.16, p=.0039), while there was no significant change in HAS (t29=-1.43, p=.16). We observed significantly more state transitions in RW compared with SD. Change in LAS and HAS across sleep conditions correlated significantly with SD vulnerability (ΔLASxΔRS: r=-0.64, p<.0001; ΔLASxΔlapse: r=0.43, p=.018; ΔHASxΔRS; r=0.43, p=.019; ΔHASxΔlapse; r=-0.39, p=.033). Finally, Δ%transitions was correlated with ΔRS but not Δlapse. Conclusion This study adds to the evidence that two specific reproducible DCS are robust markers of arousal and attention, and may be useful indicators of SD vulnerability. Support This work was supported by the National Medical Research Council, Singapore (STaR/0015/2013), and the National Research Foundation Science of Learning (NRF2016-SOL002-001).

0724 Altered Brain Network Organization in Patients With Obstructive Sleep Apnea

Abstract Introduction Previous functional MRI (fMRI) studies have reported altered brain networks in patients with obstructive sleep apnea (OSA), but the extent of such abnormal connectivity was inconsistent across studies. Moreover, despite the important role of the cerebellum in respiration and OSA, connections of the cerebellum to the cerebral cortex have been rarely assessed. Here, we investigated functional network changes in cerebral and cerebellar cortices of OSA patients. Methods Resting-state fMRI, polysomnography and neuropsychological (NP) tests data were acquired from 74 treatment naïve OSA patients (age: 45.8±10.7 years, apnea-hypopnea index: 46.4±18.5 /h) and 33 normal controls (39.6±9.3 years). Connectivity matrices were extracted by computing correlation coefficients from various ROIs, and Fisher r-to-z transformations. ROIs consisted of 234 regions matched to 17 functional networks, including 200 parcels of the cortex, and 34 parcels of the cerebellum. Between-group connectivity with age as a covariate was analyzed, and threshold for FDR correction was set at q<0.05. In the functional connections that showed the significant group differences, linear regression was conducted to examine the association between connectivity and composite score of NP tests in OSA patients. Results OSA subjects showed decreased attention, executive function, verbal fluency and verbal memory compared to controls. Resting-state functional connectivity was increased between regions involved in the default mode network (DMN), including left medial prefrontal, ventrolateral prefrontal and lateral temporal cortices. In OSA, the connectivity changes between these DMN areas negatively correlated with attention/executive function and verbal fluency. Multiple cerebellar regions showed reduces in connectivity with cerebral cortical areas including frontal eye field, temporoparietal junction, temporo-occipital gyrus, and parieto-occipital association cortex. Conclusion OSA affects mainly the DMN and cerebello-cerebral pathway. The disruption of function in these two networks are known to relate to sleep deprivation and respiratory abnormality. The abnormal DMN found in OSA patients further related to their cognitive impairment. Support This research was supported by Basic Science Research Program through the National Research Foundation of Korea funded by the Ministry of Science, ICT & Future Planning, Republic of Korea (2017R1A2B4003120) and by Samsung Biomedical Research Institute grant (OTC1190671)

Generation of Scale-Invariant Sequential Activity in Linear Recurrent Networks.

Sequential neural activity has been observed in many parts of the brain and has been proposed as a neural mechanism for memory. The natural world expresses temporal relationships at a wide range of scales. Because we cannot know the relevant scales a priori, it is desirable that memory, and thus the generated sequences, is scale invariant. Although recurrent neural network models have been proposed as a mechanism for generating sequences, the requirements for scale-invariant sequences are not known. This letter reports the constraints that enable a linear recurrent neural network model to generate scale-invariant sequential activity. A straightforward eigendecomposition analysis results in two independent conditions that are required for scale invariance for connectivity matrices with real, distinct eigenvalues. First, the eigenvalues of the network must be geometrically spaced. Second, the eigenvectors must be related to one another via translation. These constraints are easily generalizable for matrices that have complex and distinct eigenvalues. Analogous albeit less compact constraints hold for matrices with degenerate eigenvalues. These constraints, along with considerations on initial conditions, provide a general recipe to build linear recurrent neural networks that support scale-invariant sequential activity.

O2.4. FUNCTIONAL CONNECTOME-WIDE ASSOCIATIONS OF SCHIZOPHRENIA POLYGENIC RISK

BackgroundThe relationships between schizophrenia polygenic risk and connectome-wide brain connectivity remain unclear. In particular, it is unknown whether and how schizophrenia polygenic risk would influence functional connectivity both at the connectome level and in a state-independent way. In this study, we used multi-paradigm fMRI data from two independent cohorts to investigate these questions.MethodsThe discovery sample included 623 healthy Caucasian participants (age 28.86 ± 3.63 years, 302 males) acquired from the Human Connectome Project (HCP). All subjects completed fMRI scans for a battery of eight paradigms and had imputed genetic data available. Following the procedures in our prior studies, we constructed whole-brain connectivity matrices for each paradigm in each individual. From these derived matrices, we computed cross-paradigm connectivity (CPC) using principal component analysis. These CPC matrices quantify shared connectivity patterns across all paradigms for each individual and thus represent state-independent “trait” network architecture of each subject.The polygenic risk scores (PRSs) for each subject were calculated based on the genome-wide association study (GWAS) results from the Psychiatric Genomics Consortium. The scores were calculated as the sum of genome-wide risk alleles for each individual, weighted by the corresponding odds ratios to schizophrenia. We report our main findings based on the GWAS-significant threshold (P = 5×10–8). In addition, to test the robustness of our findings, we also calculated PRSs with a set of other thresholds ranging from 5×10–7 to 5×10–2.The network-based statistic (NBS) analysis was performed to associate PRSs with CPC matrices, where age, sex, and head motion were included as covariates. Significance was determined by 10,000 permutations of the original sample.The validation sample included 44 patients with schizophrenia, 43 patients with bipolar disorders, 34 patients with attention deficit hyperactivity disorder, and 77 healthy controls drawn from the Consortium for Neuropsychiatric Phenomics (CNP). All subjects completed a battery of seven fMRI paradigms. We used this sample to examine 1) whether the identified connectomic findings were specifically detected in patients with schizophrenia; and 2) whether these findings could be related to behavioral deficits in patients with schizophrenia.ResultsIn the HCP sample, the NBS analysis revealed a significant association (PFWE < 0.05) between schizophrenia PRS and a large-scale network involving a total of 69 edges connecting between 54 nodes. These nodes were predominantly distributed in the brain’s visual system, default-mode system, and frontoparietal system. Specifically, higher PRSs were associated with lower connectivity for all connections in the identified network (R = -0.37). The results were significant across all paradigms (R < -0.13, P < 0.001) and remained robust across multiple PRS thresholds (R < -0.10, P < 0.02).In the CNP sample, the connectivity of the detected network differed significantly between groups (P = 0.005), which was particularly driven by decreased connectivity in patients with SZ compared with that in HCs (PBonferroni = 0.03). The connectivity of the identified network was significantly correlated with both performance IQ (R = 0.28, P =0.002) and verbal IQ (R = 0.29, P = 0.001).DiscussionThese findings provide the first evidence for state-independent connectome-wide associations of schizophrenia polygenic risk at the systems level and suggest that disrupted integration of sensori-cognitive information may be a hallmark of genetic effects on the brain that contributes to the pathogenesis of schizophrenia.

Resting state functional network switching rate is differently altered in bipolar disorder and major depressive disorder.

The clinical misdiagnosis ratio of bipolar disorder (BD) patients to major depressive disorder (MDD) patients is high. Recent findings hypothesize that the ability to flexibly recruit functional neural networks is differently altered in BD and MDD patients. This study aimed to explore distinct aberrance of network flexibility during dynamic networks configuration in BD and MDD patients. Resting state functional magnetic resonance imaging of 40 BD patients, 61 MDD patients, and 61 matched healthy controls were recruited. Dynamic functional connectivity matrices for each subject were constructed with a sliding window method. Then, network switching rate of each node was calculated and compared among the three groups. BD and MDD patients shared decreased network switching rate of regions including left precuneus, bilateral parahippocampal gyrus, and bilateral dorsal medial prefrontal cortex. Apart from these regions, MDD patients presented specially decreased network switching rate in the bilateral anterior insula, left amygdala, and left striatum. Taken together, BD and MDD patients shared decreased network switching rate of key hubs in default mode network and MDD patients presented specially decreased switching rate in salience network and striatum. We found shared and distinct aberrance of network flexibility which revealed altered adaptive functions during dynamic networks configuration of BD and MDD.

Generalized reliability based on distances.

The intraclass correlation coefficient (ICC) is a classical index of measurement reliability. With the advent of new and complex types of data for which the ICC is not defined, there is a need for new ways to assess reliability. To meet this need, we propose a new distance‐based ICC (dbICC), defined in terms of arbitrary distances among observations. We introduce a bias correction to improve the coverage of bootstrap confidence intervals for the dbICC, and demonstrate its efficacy via simulation. We illustrate the proposed method by analyzing the test‐retest reliability of brain connectivity matrices derived from a set of repeated functional magnetic resonance imaging scans. The Spearman‐Brown formula, which shows how more intensive measurement increases reliability, is extended to encompass the dbICC.

Iterative consensus spectral clustering improves detection of subject and group level brain functional modules

Specialized processing in the brain is performed by multiple groups of brain regions organized as functional modules. Although, in vivo studies of brain functional modules involve multiple functional Magnetic Resonance Imaging (fMRI) scans, the methods used to derive functional modules from functional networks of the brain ignore individual differences in the functional architecture and use incomplete functional connectivity information. To correct this, we propose an Iterative Consensus Spectral Clustering (ICSC) algorithm that detects the most representative modules from individual dense weighted connectivity matrices derived from multiple scans. The ICSC algorithm derives group-level modules from modules of multiple individuals by iteratively minimizing the consensus-cost between the two. We demonstrate that the ICSC algorithm can be used to derive biologically plausible group-level (for multiple subjects) and subject-level (for multiple subject scans) brain modules, using resting-state fMRI scans of 589 subjects from the Human Connectome Project. We employed a multipronged strategy to show the validity of the modularizations obtained from the ICSC algorithm. We show a heterogeneous variability in the modular structure across subjects where modules involved in visual and motor processing were highly stable across subjects. Conversely, we found a lower variability across scans of the same subject. The performance of our algorithm was compared with existing functional brain modularization methods and we show that our method detects group-level modules that are more representative of the modules of multiple individuals. Finally, the experiments on synthetic images quantitatively demonstrate that the ICSC algorithm detects group-level and subject-level modules accurately under varied conditions. Therefore, besides identifying functional modules for a population of subjects, the proposed method can be used for applications in personalized neuroscience. The ICSC implementation is available at https://github.com/SCSE-Biomedical-Computing-Group/ICSC.

Cross-paradigm connectivity: reliability, stability, and utility

While functional neuroimaging studies typically focus on a particular paradigm to investigate network connectivity, the human brain appears to possess an intrinsic "trait" architecture that is independent of any given paradigm. We have previously proposed the use of "cross-paradigm connectivity (CPC)" to quantify shared connectivity patterns across multiple paradigms and have demonstrated the utility of such measures in clinical studies. Here, using generalizability theory and connectome fingerprinting, we examined the reliability, stability, and individual identifiability of CPC in a group of highly-sampled healthy traveling subjects who received fMRI scans with a battery of five paradigms across multiple sites and days. Compared with single-paradigm connectivity matrices, the CPC matrices showed higher reliability in connectivity diversity, lower reliability in connectivity strength, higher stability, and higher individual identification accuracy. All of these assessments increased as a function of number of paradigms included in the CPC analysis. In comparisons involving different paradigm combinations and different brain atlases, we observed significantly higher reliability, stability, and identifiability for CPC matrices constructed from task-only data (versus those from both task and rest data), and higher identifiability but lower stability for CPC matrices constructed from the Power atlas (versus those from the AAL atlas). Moreover, we showed that multi-paradigm CPC matrices likely reflect the brain's "trait" structure that cannot be fully achieved from single-paradigm data, even with multiple runs. The present results provide evidence for the feasibility and utility of CPC in the study of functional "trait" networks and offer some methodological implications for future CPC studies.

Working memory load-dependent changes in cortical network connectivity estimated by machine learning

Working memory engages multiple distributed brain networks to support goal-directed behavior and higher order cognition. Dysfunction in working memory has been associated with cognitive impairment in neuropsychiatric disorders. It is important to characterize the interactions among cortical networks that are sensitive to working memory load since such interactions can also hint at the impaired dynamics in patients with poor working memory performance. Functional connectivity is a powerful tool used to investigate coordinated activity among local and distant brain regions. Here, we identified connectivity footprints that differentiate task states representing distinct working memory load levels. We employed linear support vector machines to decode working memory load from task-based functional connectivity matrices in 177 healthy adults. Using neighborhood component analysis, we also identified the most important connectivity pairs in classifying high and low working memory loads. We found that between-network coupling among frontoparietal, ventral attention and default mode networks, and within-network connectivity in ventral attention network are the most important factors in classifying low vs. high working memory load. Task-based within-network connectivity profiles at high working memory load in ventral attention and default mode networks were the most predictive of load-related increases in response times. Our findings reveal the large-scale impact of working memory load on the cerebral cortex and highlight the complex dynamics of intrinsic brain networks during active task states.

On the discrete version of the black hole solution

A Schwarzschild-type solution in Regge calculus is considered. Earlier, we considered a mechanism of loose fixing of edge lengths due to the functional integral measure arising from integration over connection in the functional integral for the connection representation of the Regge action. The length scale depends on a free dimensionless parameter that determines the final functional measure. For this parameter and the length scale large in Planck units, the resulting effective action is close to the Regge action. Earlier, we considered the Regge action in terms of affine connection matrices as functions of the metric inside the 4-simplices and found that it is a finite-difference form of the Hilbert–Einstein action in the leading order over metric variations between the 4-simplices. Now we take the (continuum) Schwarzschild problem in the form where spherical symmetry is not set a priori and arises just in the solution, take the finite-difference form of the corresponding equations and get the metric (in fact, in the Lemaitre or Painlevé–Gullstrand like frame), which is nonsingular at the origin, just as the Newtonian gravitational potential, obeying the difference Poisson equation with a point source, is cutoff at the elementary length and is finite at the source.

Alterations of neural network organization during REM sleep in women: implication for sex differences in vulnerability to mood disorders

BackgroundSleep plays an important role in vulnerability to mood disorders. However, despite the existence of sex differences in vulnerability to mood disorders, no study has yet investigated the sex effect on sleep network organization and its potential involvement in vulnerability to mood disorders. The aim of our study was to empirically investigate the sex effect on network organization during REM and slow-wave sleep using the effective connectivity measured by Granger causality.MethodsPolysomnographic data from 44 healthy individuals (28 men and 16 women) recruited prospectively were analysed. To obtain the 19 × 19 connectivity matrix of all possible pairwise combinations of electrodes by Granger causality method from our EEG data, we used the Toolbox MVGC multivariate Granger causality. The computation of the network measures was realized by importing these connectivity matrices into EEGNET Toolbox.ResultsIn men and women, all small-world coefficients obtained are compatible with a small-world network organization during REM and slow-wave sleep. However, compared to men, women present greater small-world coefficients during REM sleep as well as for all EEG bands during this sleep stage, which indicates the presence of a small-world network organization less marked during REM sleep as well as for all EEG bands during this sleep stage in women. In addition, in women, these small-world coefficients during REM sleep as well as for all EEG bands during this sleep stage are positively correlated with the presence of subclinical symptoms of depression.ConclusionsThus, the highlighting of these sex differences in network organization during REM sleep indicates the presence of differences in the global and local processing of information during sleep between women and men. In addition, this small-world network organization less marked during REM sleep appears to be a marker of vulnerability to mood disorders specific to women, which opens up new perspectives in understanding sex differences in the occurrence of mood disorders.

Contextual Integration in Cortical and Convolutional Neural Networks

It has been suggested that neurons can represent sensory input using probability distributions and neural circuits can perform probabilistic inference. Lateral connections between neurons have been shown to have non-random connectivity and modulate responses to stimuli within the classical receptive field. Large-scale efforts mapping local cortical connectivity describe cell type specific connections from inhibitory neurons and like-to-like connectivity between excitatory neurons. To relate the observed connectivity to computations, we propose a neuronal network model that approximates Bayesian inference of the probability of different features being present at different image locations. We show that the lateral connections between excitatory neurons in a circuit implementing contextual integration in this should depend on correlations between unit activities, minus a global inhibitory drive. The model naturally suggests the need for two types of inhibitory gates (normalization, surround inhibition). First, using natural scene statistics and classical receptive fields corresponding to simple cells parameterized with data from mouse primary visual cortex, we show that the predicted connectivity qualitatively matches with that measured in mouse cortex: neurons with similar orientation tuning have stronger connectivity, and both excitatory and inhibitory connectivity have a modest spatial extent, comparable to that observed in mouse visual cortex. We incorporate lateral connections learned using this model into convolutional neural networks. Features are defined by supervised learning on the task, and the lateral connections provide an unsupervised learning of feature context in multiple layers. Since the lateral connections provide contextual information when the feedforward input is locally corrupted, we show that incorporating such lateral connections into convolutional neural networks makes them more robust to noise and leads to better performance on noisy versions of the MNIST dataset. Decomposing the predicted lateral connectivity matrices into low-rank and sparse components introduces additional cell types into these networks. We explore effects of cell-type specific perturbations on network computation. Our framework can potentially be applied to networks trained on other tasks, with the learned lateral connections aiding computations implemented by feedforward connections when the input is unreliable and demonstrate the potential usefulness of combining supervised and unsupervised learning techniques in real-world vision tasks.