A network engineering perspective on probing and perturbing cognition with neurofeedback

Altered controllability of white matter networks and related brain function changes in first-episode drug-naive schizophrenia

Understanding how structural connectivity alterations affect aberrant dynamic function using network control theory will provide new mechanistic insights into the pathophysiology of schizophrenia. The study included 140 drug-naive schizophrenia patients and 119 healthy controls (HCs). The average controllability (AC) quantifying capacity of brain regions/networks to shift the system into easy-to-reach states was calculated based on white matter connectivity and was compared between patients and HCs as well as functional network topological and dynamic properties. The correlation analysis between AC and duration of untreated psychosis (DUP) were conducted to characterize the controllability progression pattern without treatment effects. Relative to HCs, patients exhibited reduced AC in multiple nodes, mainly distributed in default mode network (DMN), visual network (VN), and subcortical regions, and increased AC in somatomotor network. These networks also had impaired functional topology and increased temporal variability in dynamic functional connectivity analysis. Longer DUP was related to greater reductions of AC in VN and DMN. The current study highlighted potential structural substrates underlying altered functional dynamics in schizophrenia, providing a novel understanding of the relationship of anatomic and functional network alterations.

Neurofeedback through the lens of reinforcement learning

Despite decades of experimental and clinical practice, the neuropsychological mechanisms underlying neurofeedback (NF) training remain obscure. NF is a unique form of reinforcement learning (RL) task, during which participants are provided with rewarding feedback regarding desired changes in neural patterns. However, key RL considerations - including choices during practice, prediction errors, credit-assignment problems, or the exploration-exploitation tradeoff - have infrequently been considered in the context of NF. We offer an RL-based framework for NF, describing different internal states, actions, and rewards in common NF protocols, thus fashioning new proposals for characterizing, predicting, and hastening the course of learning. In this way we hope to advance current understanding of neural regulation via NF, and ultimately to promote its effectiveness, personalization, and clinical utility.

Closed-loop Modulation of the Self-regulating Brain: A Review on Approaches, Emerging Paradigms, and Experimental Designs

Closed-loop approaches, setups, and experimental designs have been applied within the field of neuroscience to enhance the understanding of basic neurophysiology principles (closed-loop neuroscience; CLNS) and to develop improved procedures for modulating brain circuits and networks for clinical purposes (closed-loop neuromodulation; CLNM). The contents of this review are thus arranged into the following sections. First, we describe basic research findings that have been made using CLNS. Next, we provide an overview of the application, rationale, and therapeutic aspects of CLNM for clinical purposes. Finally, we summarize methodological concerns and critics in clinical practice of neurofeedback and novel applications of closed-loop perspective and techniques to improve and optimize its experiments. Moreover, we outline the theoretical explanations and experimental ideas to test animal models of neurofeedback and discuss technical issues and challenges associated with implementing closed-loop systems. We hope this review is helpful for both basic neuroscientists and clinical/ translationally-oriented scientists interested in applying closed-loop methods to improve mental health and well-being.

The impact of neurofeedback on effective connectivity networks in chronic stroke patients: an exploratory study

Objective.In this study, we assessed the impact of electroencephalography-functional magnetic resonance imaging (EEG-fMRI) neurofeedback (NF) on connectivity strength and direction in bilateral motor cortices in chronic stroke patients. Most of the studies using NF or brain computer interfaces for stroke rehabilitation have assessed treatment effects focusing on successful activation of targeted cortical regions. However, given the crucial role of brain network reorganization for stroke recovery, our broader aim was to assess connectivity changes after an NF training protocol targeting localized motor areas.Approach.We considered changes in fMRI connectivity after a multisession EEG-fMRI NF training targeting ipsilesional motor areas in nine stroke patients. We applied the dynamic causal modeling and parametric empirical Bayes frameworks for the estimation of effective connectivity changes. We considered a motor network including both ipsilesional and contralesional premotor, supplementary and primary motor areas.Main results.Our results indicate that NF upregulation of targeted areas (ipsilesional supplementary and primary motor areas) not only modulated activation patterns, but also had a more widespread impact on fMRI bilateral motor networks. In particular, inter-hemispheric connectivity between premotor and primary motor regions decreased, and ipsilesional self-inhibitory connections were reduced in strength, indicating an increase in activation during the NF motor task.Significance.To the best of our knowledge, this is the first work that investigates fMRI connectivity changes elicited by training of localized motor targets in stroke. Our results open new perspectives in the understanding of large-scale effects of NF training and the design of more effective NF strategies, based on the pathophysiology underlying stroke-induced deficits.

Improving Clinical, Cognitive, and Psychosocial Dysfunctions in Patients with Schizophrenia: A Neurofeedback Randomized Control Trial

Objectives

The aim of this study was to use neurofeedback (NF) training as the add-on therapy in patients with schizophrenia to improve their clinical, cognitive, and psychosocial condition. The study, thanks to the monitoring of various conditions, quantitative electroencephalogram (QEEG) and brain-derived neurotrophic factor (BDNF), was supposed to give an insight into mechanisms underlying NF training results.

Methods

Forty-four male patients with schizophrenia, currently in a stable, incomplete remission, were recruited into two, 3-month rehabilitation programs, with standard rehabilitation as a control group (R) or with add-on NF training (NF). Pre- and posttherapy primary outcomes were compared: clinical (Positive and Negative Syndrome Scale (PANSS)), cognitive (Color Trails Test (CTT), d2 test), psychosocial functioning (General Self-Efficacy Scale (GSES), Beck Cognitive Insight Scale (BCIS), and Acceptance of Illness Scale (AIS)), quantitative electroencephalogram (QEEG), auditory event-related potentials (ERPs), and serum level of BDNF. Results. Both groups R and NF improved significantly in clinical ratings (Positive and Negative Syndrome Scale (PANSS)). In-between analyses unveiled some advantages of add-on NF therapy over standard rehabilitation. GSES scores improved significantly, giving the NF group of patients greater ability to cope with stressful or difficult social demands. Also, the serum-level BDNF increased significantly more in the NF group. Post hoc analyses indicated the possibility of creating a separate PANSS subsyndrome, specifically related to cognitive, psychosocial, and BDNF effects of NF therapy.

Conclusions

Neurofeedback can be effectively used as the add-on therapy in schizophrenia rehabilitation programs. The method requires further research regarding its clinical specificity and understanding mechanisms of action.

Predictors of real-time fMRI neurofeedback performance and improvement - A machine learning mega-analysis

Real-time fMRI neurofeedback is an increasingly popular neuroimaging technique that allows an individual to gain control over his/her own brain signals, which can lead to improvements in behavior in healthy participants as well as to improvements of clinical symptoms in patient populations. However, a considerably large ratio of participants undergoing neurofeedback training do not learn to control their own brain signals and, consequently, do not benefit from neurofeedback interventions, which limits clinical efficacy of neurofeedback interventions. As neurofeedback success varies between studies and participants, it is important to identify factors that might influence neurofeedback success. Here, for the first time, we employed a big data machine learning approach to investigate the influence of 20 different design-specific (e.g. activity vs. connectivity feedback), region of interest-specific (e.g. cortical vs. subcortical) and subject-specific factors (e.g. age) on neurofeedback performance and improvement in 608 participants from 28 independent experiments. With a classification accuracy of 60% (considerably different from chance level), we identified two factors that significantly influenced neurofeedback performance: Both the inclusion of a pre-training no-feedback run before neurofeedback training and neurofeedback training of patients as compared to healthy participants were associated with better neurofeedback performance. The positive effect of pre-training no-feedback runs on neurofeedback performance might be due to the familiarization of participants with the neurofeedback setup and the mental imagery task before neurofeedback training runs. Better performance of patients as compared to healthy participants might be driven by higher motivation of patients, higher ranges for the regulation of dysfunctional brain signals, or a more extensive piloting of clinical experimental paradigms. Due to the large heterogeneity of our dataset, these findings likely generalize across neurofeedback studies, thus providing guidance for designing more efficient neurofeedback studies specifically for improving clinical neurofeedback-based interventions. To facilitate the development of data-driven recommendations for specific design details and subpopulations the field would benefit from stronger engagement in open science research practices and data sharing.

Psychiatric Neural Networks and Precision Therapeutics by Machine Learning

Learning and environmental adaptation increase the likelihood of survival and improve the quality of life. However, it is often difficult to judge optimal behaviors in real life due to highly complex social dynamics and environment. Consequentially, many different brain regions and neuronal circuits are involved in decision-making. Many neurobiological studies on decision-making show that behaviors are chosen through coordination among multiple neural network systems, each implementing a distinct set of computational algorithms. Although these processes are commonly abnormal in neurological and psychiatric disorders, the underlying causes remain incompletely elucidated. Machine learning approaches with multidimensional data sets have the potential to not only pathologically redefine mental illnesses but also better improve therapeutic outcomes than DSM/ICD diagnoses. Furthermore, measurable endophenotypes could allow for early disease detection, prognosis, and optimal treatment regime for individuals. In this review, decision-making in real life and psychiatric disorders and the applications of machine learning in brain imaging studies on psychiatric disorders are summarized, and considerations for the future clinical translation are outlined. This review also aims to introduce clinicians, scientists, and engineers to the opportunities and challenges in bringing artificial intelligence into psychiatric practice.

Reinforcement of Self-Regulated Brain Activity in Schizophrenia Patients Undergoing Rehabilitation

The experiment was aimed to compare the effects of different forms of rehabilitation applied in patients with schizophrenia. Verification of the obtained results was based on the analysis of the level of cognitive and social functioning of the subjects. For this purpose, the following clinical tools were used: Positive and Negative Syndrome Scale (PANSS), Beck Cognitive Insight Scale (BCIS), Color Trial Test (CTT-1, CTT-2), d2 psychological tests, Acceptance of Illness Scale (AIS), Self-efficacy Scale (GSES), Quantitative Electroencephalogram Biofeedback (QEEG-BF), auditory event-related potentials (ERPs), and serum levels of brain-derived neurotrophic factor (BDNF). The subjects were mentally stable male schizophrenia patients who had been in remission. They were divided into two groups which received different types of rehabilitation for three months. Group 1 patients followed a standard rehabilitation and Group 2 patients received GSR Biofeedback (galvanic skin response Biofeedback, GSR-BF) training. Pretherapy and posttherapy measurements were made for each group. Experimental rehabilitation based on GSR-BF training resulted in regulatory control of neurophysiological mechanisms, and the parameters obtained demonstrated improvement in the subjects' cognitive and social function. The following therapy outcomes were observed: (1) reduce psychopathological symptoms (2) improving cognitive (concentration, attention) and social functions (3) increase in the neurotrophic factor BDNF. GSR-BF can be used as an alternative to conventional rehabilitation in schizophrenia patients.

Improving Emotion Regulation Through Real-Time Neurofeedback Training on the Right Dorsolateral Prefrontal Cortex: Evidence From Behavioral and Brain Network Analyses

We investigated if emotion regulation can be improved through self-regulation training on non-emotional brain regions, as well as how to change the brain networks implicated in this process. During the training period, the participants were instructed to up-regulate their right dorsolateral prefrontal cortex (rDLPFC) activity according to real-time functional near-infrared spectroscopy (fNIRS) neurofeedback signals, and there was no emotional element. The results showed that the training significantly increased emotion regulation, resting-state functional connectivity (rsFC) within the emotion regulation network (ERN) and frontoparietal network (FPN), and rsFC between the ERN and amygdala; however, training did not influence the rsFC between the FPN and the amygdala. However, self-regulation training on rDLPFC significantly improved emotion regulation and generally increased the rsFCs within the networks; the rsFC between the ERN and amygdala was also selectively increased. The present study also described a safe approach that may improve emotion regulation through self-regulation training on non-emotional brain regions.

Directly Exploring the Neural Correlates of Feedback-Related Reward Saliency and Valence During Real-Time fMRI-Based Neurofeedback

Introduction: The potential therapeutic efficacy of real-time fMRI Neurofeedback has received increasing attention in a variety of psychological and neurological disorders and as a tool to probe cognition. Despite its growing popularity, the success rate varies significantly, and the underlying neural mechanisms are still a matter of debate. The question whether an individually tailored framework positively influences neurofeedback success remains largely unexplored.

Methods: To address this question, participants were trained to modulate the activity of a target brain region, the visual motion area hMT+/V5, based on the performance of three imagery tasks with increasing complexity: imagery of a static dot, imagery of a moving dot with two and with four opposite directions. Participants received auditory feedback in the form of vocalizations with either negative, neutral or positive valence. The modulation thresholds were defined for each participant according to the maximum BOLD signal change of their target region during the localizer run.

Results: We found that 4 out of 10 participants were able to modulate brain activity in this region-of-interest during neurofeedback training. This rate of success (40%) is consistent with the neurofeedback literature. Whole-brain analysis revealed the recruitment of specific cortical regions involved in cognitive control, reward monitoring, and feedback processing during neurofeedback training. Individually tailored feedback thresholds did not correlate with the success level. We found region-dependent neuromodulation profiles associated with task complexity and feedback valence.

Discussion: Findings support the strategic role of task complexity and feedback valence on the modulation of the network nodes involved in monitoring and feedback control, key variables in neurofeedback frameworks optimization. Considering the elaborate design, the small sample size here tested (N = 10) impairs external validity in comparison to our previous studies. Future work will address this limitation. Ultimately, our results contribute to the discussion of individually tailored solutions, and justify further investigation concerning volitional control over brain activity.

On the Nature of Explanations Offered by Network Science: A Perspective From and for Practicing Neuroscientists

Network neuroscience represents the brain as a collection of regions and inter-regional connections. Given its ability to formalize systems-level models, network neuroscience has generated unique explanations of neural function and behavior. The mechanistic status of these explanations and how they can contribute to and fit within the field of neuroscience as a whole has received careful treatment from philosophers. However, these philosophical contributions have not yet reached many neuroscientists. Here we complement formal philosophical efforts by providing an applied perspective from and for neuroscientists. We discuss the mechanistic status of the explanations offered by network neuroscience and how they contribute to, enhance, and interdigitate with other types of explanations in neuroscience. In doing so, we rely on philosophical work concerning the role of causality, scale, and mechanisms in scientific explanations. In particular, we make the distinction between an explanation and the evidence supporting that explanation, and we argue for a scale-free nature of mechanistic explanations. In the course of these discussions, we hope to provide a useful applied framework in which network neuroscience explanations can be exercised across scales and combined with other fields of neuroscience to gain deeper insights into the brain and behavior.

Can we predict real-time fMRI neurofeedback learning success from pretraining brain activity?

Neurofeedback training has been shown to influence behavior in healthy participants as well as to alleviate clinical symptoms in neurological, psychosomatic, and psychiatric patient populations. However, many real-time fMRI neurofeedback studies report large inter-individual differences in learning success. The factors that cause this vast variability between participants remain unknown and their identification could enhance treatment success. Thus, here we employed a meta-analytic approach including data from 24 different neurofeedback studies with a total of 401 participants, including 140 patients, to determine whether levels of activity in target brain regions during pretraining functional localizer or no-feedback runs (i.e., self-regulation in the absence of neurofeedback) could predict neurofeedback learning success. We observed a slightly positive correlation between pretraining activity levels during a functional localizer run and neurofeedback learning success, but we were not able to identify common brain-based success predictors across our diverse cohort of studies. Therefore, advances need to be made in finding robust models and measures of general neurofeedback learning, and in increasing the current study database to allow for investigating further factors that might influence neurofeedback learning.

Primary functional brain connections associated with melancholic major depressive disorder and modulation by antidepressants

The limited efficacy of available antidepressant therapies may be due to how they affect the underlying brain network. The purpose of this study was to develop a melancholic MDD biomarker to identify critically important functional connections (FCs), and explore their association to treatments. Resting state fMRI data of 130 individuals (65 melancholic major depressive disorder (MDD) patients, 65 healthy controls) were included to build a melancholic MDD classifier, and 10 FCs were selected by our sparse machine learning algorithm. This biomarker generalized to a drug-free independent cohort of melancholic MDD, and did not generalize to other MDD subtypes or other psychiatric disorders. Moreover, we found that antidepressants had a heterogeneous effect on the identified FCs of 25 melancholic MDDs. In particular, it did impact the FC between left dorsolateral prefrontal cortex (DLPFC)/inferior frontal gyrus (IFG) and posterior cingulate cortex (PCC)/precuneus, ranked as the second 'most important' FC based on the biomarker weights, whilst other eight FCs were normalized. Given that left DLPFC has been proposed as an explicit target of depression treatments, this suggest that the limited efficacy of antidepressants might be compensated by combining therapies with targeted treatment as an optimized approach in the future.

EEG Feature Extraction Based on a Bilevel Network: Minimum Spanning Tree and Regional Network

Feature extraction is essential for classifying different motor imagery (MI) tasks in a brain–computer interface (BCI). Although the methods of brain network analysis have been widely studied in the BCI field, these methods are limited by differences in network size, density, and standardization. To address this issue and improve classification accuracy, we propose a novel method, in which the hybrid features of the brain function based on the bilevel network are extracted. Minimum spanning tree (MST) based on electroencephalogram (EEG) signal nodes in different MIs is constructed as the first network layer to solve the global network connectivity problem. In addition, the regional network in different movement patterns is constructed as the second network layer to determine the network characteristics, which is consistent with the correspondence between limb movement patterns and cerebral cortex in neurophysiology. We attempt to apply MST to the classification of the MI EEG signals, and the bilevel network has better interpretability. Thereafter, a vector is formed by combining the MST fundamental features with the directional features of the regional network. Our method is validated using the BCI Competition IV Dataset I. Experimental results verify the feasibility of the bilevel network framework. Furthermore, the average classification performance of the proposed method reaches 89.50%, which is higher than that of other competing methods, thereby indicating that the bilevel network is effective for MI classification.

Functional disconnection of associative cortical areas predicts performance during BCI training

Brain-computer interfaces (BCIs) have been largely developed to allow communication, control, and neurofeedback in human beings. Despite their great potential, BCIs perform inconsistently across individuals and the neural processes that enable humans to achieve good control remain poorly understood. To address this question, we performed simultaneous high-density electroencephalographic (EEG) and magnetoencephalographic (MEG) recordings in a motor imagery-based BCI training involving a group of healthy subjects. After reconstructing the signals at the cortical level, we showed that the reinforcement of motor-related activity during the BCI skill acquisition is paralleled by a progressive disconnection of associative areas which were not directly targeted during the experiments. Notably, these network connectivity changes reflected growing automaticity associated with BCI performance and predicted future learning rate. Altogether, our findings provide new insights into the large-scale cortical organizational mechanisms underlying BCI learning, which have implications for the improvement of this technology in a broad range of real-life applications.

Network topology dynamics of circulating biomarkers and cognitive performance in older Cytomegalovirus-seropositive or -seronegative men and women

Background

Cytokines are signaling molecules operating within complex cascade patterns and having exceptional modulatory functions. They impact various physiological processes such as neuroendocrine and metabolic interactions, neurotrophins’ metabolism, neuroplasticity, and may affect behavior and cognition. In our previous study, we found that sex and Cytomegalovirus (CMV)-serostatus may modulate levels of circulating pro- and anti-inflammatory cytokines, metabolic factors, immune cells, and cognitive performance, as well as associations between them.

Results

In the present study, we used a graph-theoretical approach to investigate the network topology dynamics of 22 circulating biomarkers and 11 measures of cognitive performance in 161 older participants recruited to undergo a six-months training intervention. For network construction, we applied coefficient of determination (R²) that was calculated for all possible pairs of variables (N = 33) in four groups (CMV⁻ men and women; CMV⁺ men and women). Network topology has been evaluated by clustering coefficient (CC) and characteristic path length (CPL) as well as local (E_local) and global (E_global) efficiency, showing the degree of network segregation (CC and E_local) and integration (CPL and E_global). We found that networks under consideration showed small-world networks properties with more random characteristics. Mean CC, as well as local and global efficiency were highest and CPL shortest in CMV⁻ males (having lowest inflammatory status and highest cognitive performance). CMV⁻ and CMV⁺ females did not show any significant differences. Modularity analyses showed that the networks exhibit in all cases highly differentiated modular organization (with Q-value ranged between 0.397 and 0.453).

Conclusions

In this work, we found that segregation and integration properties of the network were notably stronger in the group with balanced inflammatory status. We were also able to confirm our previous findings that CMV-infection and sex modulate multiple circulating biomarkers and cognitive performance and that balanced inflammatory and metabolic status in elderly contributes to better cognitive functioning. Thus, network analyses provide a useful strategy for visualization and quantitative description of multiple interactions between various circulating pro- and anti-inflammatory biomarkers, hormones, neurotrophic and metabolic factors, immune cells, and measures of cognitive performance and can be in general applied for analyzing interactions between different physiological systems.

Pushing the Boundaries of Psychiatric Neuroimaging to Ground Diagnosis in Biology

To accurately detect, track progression of, and develop novel treatments for mental illnesses, a diagnostic framework is needed that is grounded in biological features. Here, we present the case for utilizing personalized neuroimaging, computational modeling, standardized computing, and ecologically valid neuroimaging to anchor psychiatric nosology in biology.

What would a synthetic connectome look like?

A major challenge of contemporary neuroscience is to unravel the structure of the connectome, the ensemble of neural connections that link between different functional units of the brain, and to reveal how this structure relates to brain function. This thriving area of research largely follows the general tradition in biology of reverse-engineering, which consists of first observing and characterizing a biological system or process, and then deconstructing it into its fundamental building blocks in order to infer its modes of operation. However, a complementary form of biology has emerged, synthetic biology, which emphasizes construction-based forward-engineering. The synthetic biology approach comprises the assembly of new biological systems out of elementary biological parts. The rationale is that the act of building a system can be a powerful method for gaining deep understanding of how that system works. As the fields of connectomics and synthetic biology are independently growing, I propose to consider the benefits of combining the two, to create synthetic connectomics, a new form of neuroscience and a new form of synthetic biology. The goal of synthetic connectomics would be to artificially design and construct the connectomes of live behaving organisms. Synthetic connectomics could serve as a unifying platform for unraveling the complexities of brain operation and perhaps also for generating new forms of artificial life, and, in general, could provide a valuable opportunity for empirically exploring theoretical predictions about network function. What would a synthetic connectome look like? What purposes would it serve? How could it be constructed? This review delineates the novel notion of a synthetic connectome and aims to lay out the initial steps towards its implementation, contemplating its impact on science and society.

Hyper-Frequency Network Topology Changes During Choral Singing

Choral singing requires the coordination of physiological subsystems within and across individuals. Previously, we suggested that the choir functions as a superordinate system that imposes boundary conditions on the dynamic features of the individual singers and found reliable differences in the network topography by analyzing within- and cross-frequency couplings (WFC and CFC, respectively). Here, we further refine our analyses to investigate hyper-frequency network (HFN) topology structures (i.e., the layout or arrangement of connections) using a graph-theoretical approach. In a sample of eleven singers and one conductor engaged in choral singing (aged between 23 and 56 years, and including five men and seven women), we calculated phase coupling (WFC and CFC) between respiratory, cardiac, and vocalizing subsystems across ten frequencies of interest. All these couplings were used for construction of HFN with nodes being a combination of frequency components and subsystems across choir participants. With regard to the network topology measures, we found that clustering coefficients (CCs) as well as local and global efficiency were highest and characteristic path lengths, correspondingly, were shortest when the choir sang a canon in parts as compared to singing it in unison. Furthermore, these metrics revealed a significant relationship to individual heart rate, as an indicator of arousal, and to an index of heart rate variability indicated by the LF/HF ratio (low and high frequency, respectively), and reflecting the balance between sympathetic and parasympathetic activity. In addition, we found that the CC and local efficiency for groups singing the same canon part were higher than for groups of singers constructed randomly post hoc, indicating stronger neighbor–neighbor connections in the former. We conclude that network topology dynamics are a crucial determinant of group behavior and may represent a potent biomarker for social interaction.

A framework for offline evaluation and optimization of real-time algorithms for use in neurofeedback, demonstrated on an instantaneous proxy for correlations

Interest in real-time fMRI neurofeedback has grown exponentially over the past few years, both for use as a basic science research tool, and as part of the search for novel clinical interventions for neurological and psychiatric illnesses. In order to expand the range of questions which can be addressed with this tool however, new neurofeedback methods must be developed, going beyond feedback of activations in a single region. These new methods, several of which have already been proposed, are by their nature complex, involving many possible parameters. Here we suggest a framework for evaluating and optimizing algorithms for use in a real-time setting, before beginning the neurofeedback experiment, by offline simulations of algorithm output using a previously collected dataset. We demonstrate the application of this framework on the instantaneous proxy for correlations which we developed for training connectivity between different network nodes, identify the optimal parameters for use with this algorithm, and compare it to more traditional correlation methods. We also examine the effects of advanced imaging techniques, such as multi-echo acquisition, and the integration of these into the real-time processing stream.

Network neuroscience for optimizing brain-computer interfaces

Human-machine interactions are being increasingly explored to create alternative ways of communication and to improve our daily life. Based on a classification of the user's intention from the user's underlying neural activity, brain-computer interfaces (BCIs) allow direct interactions with the external environment while bypassing the traditional effector of the musculoskeletal system. Despite the enormous potential of BCIs, there are still a number of challenges that limit their societal impact, ranging from the correct decoding of a human's thoughts, to the application of effective learning strategies. Despite several important engineering advances, the basic neuroscience behind these challenges remains poorly explored. Indeed, BCIs involve complex dynamic changes related to neural plasticity at a diverse range of spatiotemporal scales. One promising antidote to this complexity lies in network science, which provides a natural language in which to model the organizational principles of brain architecture and function as manifest in its interconnectivity. Here, we briefly review the main limitations currently affecting BCIs, and we offer our perspective on how they can be addressed by means of network theoretic approaches. We posit that the emerging field of network neuroscience will prove to be an effective tool to unlock human-machine interactions.

Dynamic graph metrics: Tutorial, toolbox, and tale

The central nervous system is composed of many individual units - from cells to areas - that are connected with one another in a complex pattern of functional interactions that supports perception, action, and cognition. One natural and parsimonious representation of such a system is a graph in which nodes (units) are connected by edges (interactions). While applicable across spatiotemporal scales, species, and cohorts, the traditional graph approach is unable to address the complexity of time-varying connectivity patterns that may be critically important for an understanding of emotional and cognitive state, task-switching, adaptation and development, or aging and disease progression. Here we survey a set of tools from applied mathematics that offer measures to characterize dynamic graphs. Along with this survey, we offer suggestions for visualization and a publicly-available MATLAB toolbox to facilitate the application of these metrics to existing or yet-to-be acquired neuroimaging data. We illustrate the toolbox by applying it to a previously published data set of time-varying functional graphs, but note that the tools can also be applied to time-varying structural graphs or to other sorts of relational data entirely. Our aim is to provide the neuroimaging community with a useful set of tools, and an intuition regarding how to use them, for addressing emerging questions that hinge on accurate and creative analyses of dynamic graphs.

Dynamic graph metrics: Tutorial, toolbox, and tale

The central nervous system is composed of many individual units - from cells to areas - that are connected with one another in a complex pattern of functional interactions that supports perception, action, and cognition. One natural and parsimonious representation of such a system is a graph in which nodes (units) are connected by edges (interactions). While applicable across spatiotemporal scales, species, and cohorts, the traditional graph approach is unable to address the complexity of time-varying connectivity patterns that may be critically important for an understanding of emotional and cognitive state, task-switching, adaptation and development, or aging and disease progression. Here we survey a set of tools from applied mathematics that offer measures to characterize dynamic graphs. Along with this survey, we offer suggestions for visualization and a publicly-available MATLAB toolbox to facilitate the application of these metrics to existing or yet-to-be acquired neuroimaging data. We illustrate the toolbox by applying it to a previously published data set of time-varying functional graphs, but note that the tools can also be applied to time-varying structural graphs or to other sorts of relational data entirely. Our aim is to provide the neuroimaging community with a useful set of tools, and an intuition regarding how to use them, for addressing emerging questions that hinge on accurate and creative analyses of dynamic graphs.

Understanding the Emergence of Neuropsychiatric Disorders With Network Neuroscience

Major neuropsychiatric disorders such as psychosis are increasingly acknowledged to be disorders of brain connectivity. Yet tools to map, model, predict, and change connectivity are difficult to develop, largely because of the complex, dynamic, and multivariate nature of interactions between brain regions. Network neuroscience (NN) provides a theoretical framework and mathematical toolset to address these difficulties. Building on areas of mathematics such as graph theory, NN in its simplest form summarizes neuroimaging data by treating brain regions as nodes in a graph and by treating interactions or connections between nodes as edges in the graph. Network metrics can then be used to quantitatively describe the architecture of the graph, which in turn reflects the network's function. We review evidence supporting the utility of NN in understanding psychiatric disorders, with a focus on normative brain network development and abnormalities associated with psychosis. We also emphasize relevant methodological challenges, such as motion artifact correction, which are particularly important to consider when applying network tools to developmental neuroimaging data. We close with a discussion of several emerging frontiers of NN in psychiatry, including generative network modeling and network control theory. We aim to offer an accessible introduction to this emerging field and motivate further work that uses NN to better understand the normative development of brain networks and alterations in that development that accompany or foreshadow psychiatric disease.

Large-scale dynamic modeling of task-fMRI signals via subspace system identification

OBJECTIVE

We analyze task-based fMRI time series to produce large-scale dynamical models that are capable of approximating the observed signal with good accuracy.

APPROACH

We extend subspace system identification methods for deterministic and stochastic state-space models with external inputs. The dynamic behavior of the generated models is characterized using control-theoretic analysis tools. To validate their effectiveness, we perform a probabilistic inversion of the identified input-output relationships via joint state-input maximum likelihood estimation. Our experimental setup explores a large dataset generated using state-of-the-art acquisition and pre-processing methods from the Human Connectome Project.

MAIN RESULTS

We analyze both anatomically parcellated and spatially dense time series, and propose an efficient algorithm to address the high-dimensional optimization problem resulting from the latter. Our results enable the quantification of input-output transfer functions between each task condition and each region of the cortex, as exemplified by a motor task. Further, the identified models produce impulse response functions between task conditions and cortical regions that are compatible with typical hemodynamic response functions. We then extend subspace methods to account for multi-subject experimental configurations, identifying models that capture common dynamical characteristics across subjects. Finally, we show that system inversion via maximum-likelihood allows the time-of-occurrence of the task stimuli to be estimated from the observed outputs.

SIGNIFICANCE

The ability to produce dynamical input-output models might have an impact in the expanding field of neurofeedback. In particular, the models we produce allow the partial quantification of the effect of external task-related inputs on the metabolic response of the brain, conditioned on its current state. Such a notion provides a basis for leveraging control-theoretic approaches to neuromodulation and self-regulation in therapeutic applications.

Modulating functional connectivity after stroke with neurofeedback: Effect on motor deficits in a controlled cross-over study

Synchronization of neural activity as measured with functional connectivity (FC) is increasingly used to study the neural basis of brain disease and to develop new treatment targets. However, solid evidence for a causal role of FC in disease and therapy is lacking. Here, we manipulated FC of the ipsilesional primary motor cortex in ten chronic human stroke patients through brain-computer interface technology with visual neurofeedback. We conducted a double-blind controlled crossover study to test whether manipulation of FC through neurofeedback had a behavioral effect on motor performance. Patients succeeded in increasing FC in the motor cortex. This led to improvement in motor function that was significantly greater than during neurofeedback training of a control brain area and proportional to the degree of FC enhancement. This result provides evidence that FC has a causal role in neurological function and that it can be effectively targeted with therapy.

•

Stroke patients participated in clinical trial on neurofeedback of functional connectivity.

•

Patients learned to enhance synchrony of neural activity in their motor cortex.

•

This led to reduced motor impairment.

•

Evidence for a causal role of neural synchrony in neurological deficits and recovery.

Subgraphs of functional brain networks identify dynamical constraints of cognitive control

Brain anatomy and physiology support the human ability to navigate a complex space of perceptions and actions. To maneuver across an ever-changing landscape of mental states, the brain invokes cognitive control-a set of dynamic processes that engage and disengage different groups of brain regions to modulate attention, switch between tasks, and inhibit prepotent responses. Current theory posits that correlated and anticorrelated brain activity may signify cooperative and competitive interactions between brain areas that subserve adaptive behavior. In this study, we use a quantitative approach to identify distinct topological motifs of functional interactions and examine how their expression relates to cognitive control processes and behavior. In particular, we acquire fMRI BOLD signal in twenty-eight healthy subjects as they perform two cognitive control tasks-a Stroop interference task and a local-global perception switching task using Navon figures-each with low and high cognitive control demand conditions. Based on these data, we construct dynamic functional brain networks and use a parts-based, network decomposition technique called non-negative matrix factorization to identify putative cognitive control subgraphs whose temporal expression captures distributed network structures involved in different phases of cooperative and competitive control processes. Our results demonstrate that temporal expression of the subgraphs fluctuate alongside changes in cognitive demand and are associated with individual differences in task performance. These findings offer insight into how coordinated changes in the cooperative and competitive roles of cognitive systems map trajectories between cognitively demanding brain states.

Predicting future learning from baseline network architecture

Human behavior and cognition result from a complex pattern of interactions between brain regions. The flexible reconfiguration of these patterns enables behavioral adaptation, such as the acquisition of a new motor skill. Yet, the degree to which these reconfigurations depend on the brain's baseline sensorimotor integration is far from understood. Here, we asked whether spontaneous fluctuations in sensorimotor networks at baseline were predictive of individual differences in future learning. We analyzed functional MRI data from 19 participants prior to six weeks of training on a new motor skill. We found that visual-motor connectivity was inversely related to learning rate: sensorimotor autonomy at baseline corresponded to faster learning in the future. Using three additional scans, we found that visual-motor connectivity at baseline is a relatively stable individual trait. These results suggest that individual differences in motor skill learning can be predicted from sensorimotor autonomy at baseline prior to task execution.

Fronto-limbic dysconnectivity leads to impaired brain network controllability in young people with bipolar disorder and those at high genetic risk

Recent investigations have used diffusion-weighted imaging to reveal disturbances in the neurocircuitry that underlie cognitive-emotional control in bipolar disorder (BD) and in unaffected siblings or children at high genetic risk (HR). It has been difficult to quantify the mechanism by which structural changes disrupt the superimposed brain dynamics, leading to the emotional lability that is characteristic of BD. Average controllability is a concept from network control theory that extends structural connectivity data to estimate the manner in which local neuronal fluctuations spread from a node or subnetwork to alter the state of the rest of the brain. We used this theory to ask whether structural connectivity deficits previously observed in HR individuals (n = 84, mean age 22.4), patients with BD (n = 38, mean age 23.9), and age- and gender-matched controls (n = 96, mean age 22.6) translate to differences in the ability of brain systems to be manipulated between states. Localized impairments in network controllability were seen in the left parahippocampal, left middle occipital, left superior frontal, right inferior frontal, and right precentral gyri in BD and HR groups. Subjects with BD had distributed deficits in a subnetwork containing the left superior and inferior frontal gyri, postcentral gyrus, and insula (p = 0.004). HR participants had controllability deficits in a right-lateralized subnetwork involving connections between the dorsomedial and ventrolateral prefrontal cortex, the superior temporal pole, putamen, and caudate nucleus (p = 0.008). Between-group controllability differences were attenuated after removal of topological factors by network randomization. Some previously reported differences in network connectivity were not associated with controllability-differences, likely reflecting the contribution of more complex brain network properties. These analyses highlight the potential functional consequences of altered brain networks in BD, and may guide future clinical interventions.

•

Control theory estimates how neuronal fluctuations spread from local networks.

•

We compare brain controllability in bipolar disorder and their high-risk relatives.

•

These groups have impaired controllability in networks supporting cognitive and emotional control.

•

Weaker connectivity as well as topological alterations contribute to these changes.

From Maps to Multi-dimensional Network Mechanisms of Mental Disorders

The development of advanced neuroimaging techniques and their deployment in large cohorts has enabled an assessment of functional and structural brain network architecture at an unprecedented level of detail. Across many temporal and spatial scales, network neuroscience has emerged as a central focus of intellectual efforts, seeking meaningful descriptions of brain networks and explanatory sets of network features that underlie circuit function in health and dysfunction in disease. However, the tools of network science commonly deployed provide insight into brain function at a fundamentally descriptive level, often failing to identify (patho-)physiological mechanisms that link system-level phenomena to the multiple hierarchies of brain function. Here we describe recently developed techniques stemming from advances in complex systems and network science that have the potential to overcome this limitation, thereby contributing mechanistic insights into neuroanatomy, functional dynamics, and pathology. Finally, we build on the Research Domain Criteria framework, highlighting the notion that mental illnesses can be conceptualized as dysfunctions of neural circuitry present across conventional diagnostic boundaries, to sketch how network-based methods can be combined with pharmacological, intermediate phenotype, genetic, and magnetic stimulation studies to probe mechanisms of psychopathology.

Generative models for network neuroscience: prospects and promise

Network neuroscience is the emerging discipline concerned with investigating the complex patterns of interconnections found in neural systems, and identifying principles with which to understand them. Within this discipline, one particularly powerful approach is network generative modelling, in which wiring rules are algorithmically implemented to produce synthetic network architectures with the same properties as observed in empirical network data. Successful models can highlight the principles by which a network is organized and potentially uncover the mechanisms by which it grows and develops. Here, we review the prospects and promise of generative models for network neuroscience. We begin with a primer on network generative models, with a discussion of compressibility and predictability, and utility in intuiting mechanisms, followed by a short history on their use in network science, broadly. We then discuss generative models in practice and application, paying particular attention to the critical need for cross-validation. Next, we review generative models of biological neural networks, both at the cellular and large-scale level, and across a variety of species including Caenorhabditis elegans, Drosophila, mouse, rat, cat, macaque and human. We offer a careful treatment of a few relevant distinctions, including differences between generative models and null models, sufficiency and redundancy, inferring and claiming mechanism, and functional and structural connectivity. We close with a discussion of future directions, outlining exciting frontiers both in empirical data collection efforts as well as in method and theory development that, together, further the utility of the generative network modelling approach for network neuroscience.

Direct modulation of aberrant brain network connectivity through real-time NeuroFeedback

The existence of abnormal connectivity patterns between resting state networks in neuropsychiatric disorders, including Autism Spectrum Disorder (ASD), has been well established. Traditional treatment methods in ASD are limited, and do not address the aberrant network structure. Using real-time fMRI neurofeedback, we directly trained three brain nodes in participants with ASD, in which the aberrant connectivity has been shown to correlate with symptom severity. Desired network connectivity patterns were reinforced in real-time, without participants’ awareness of the training taking place. This training regimen produced large, significant long-term changes in correlations at the network level, and whole brain analysis revealed that the greatest changes were focused on the areas being trained. These changes were not found in the control group. Moreover, changes in ASD resting state connectivity following the training were correlated to changes in behavior, suggesting that neurofeedback can be used to directly alter complex, clinically relevant network connectivity patterns.

Even when we are at rest, our brains are always active. For example, areas of the brain involved in vision remain active in complete darkness. Different brain regions that connect together to perform a given task often show coordinated activity at rest. Past studies have shown that these resting connections are different in people with conditions such as autism. Some brain regions are more weakly connected while others are more strongly connected in people with autism spectrum disorder compared to those without. Furthermore, people with more severe symptoms seem to have more abnormal connections.

“Neurofeedback training” is a method of changing the resting connections between different brain regions. Scientists measure a brain signal – the connection between different brain regions – from a person in real time. They then provide positive feedback to the person if this signal improves. For example, if a connection that is too weak becomes stronger, the scientists might reinforce this by providing feedback on the success. Previous work has shown that neurofeedback training may even change people’s behaviour. However, it has not yet been explored as a method of treating the abnormal connections seen in people with autism when their brains are at rest.

To address this, Ramot et al. used a technique known as “functional magnetic resonance imaging” (or fMRI for short) to measure brain activity in young men with autism. First, certain brain regions were identified as having abnormal resting connections with each other. The participants were then asked to look at a blank screen and to try to reveal a picture hidden underneath. Whenever the connections between the chosen brain regions improved, part of the picture was revealed on the screen, accompanied by an upbeat sound. The participants were unaware that it was their brain signals causing this positive feedback.

This form of neurofeedback training successfully changed the abnormal brain connections in most of the participants with autism, making their connections more similar to those seen in the wider population. These effects lasted up to a year after training. Early results also suggest that these changes were related to improvements in symptoms, although further work is needed to see if doctors could reliably use this method as a therapy. These findings show that neurofeedback training could potentially help treat not only autism spectrum disorder, but a range of other disorders that involve abnormal brain connections, including depression and schizophrenia.