ShiNung Ching, Ph.D. - US grants

The past decade has seen significant growth in the development and use of neurostimulation technology to manipulate neural activity in the brain. The applications of such technology range from scientific objectives, e.g., studying how different parts of the brain interact with each other, to clinical objectives, e.g., using stimulation to alleviate the symptoms of neurological disorders such as Parkinson's disease. Despite many technological advances associated with such stimulation, its use is still largely limited to perturbative paradigms, in which stereotyped inputs (waveforms) are used to activate or deactivate a neuronal network in its entirety. In other words, the stimulation is used to create a uniform circuit response, turning an entire population of brain cells (neurons) on or off, without regard for specificity (i.e., which cells in the population respond) or timing (i.e., when they turn on or off). In engineering, control is understood as not simply uniform stimulation, but as the precise creation or prevention of certain system maneuvers at each moment in time. This project will investigate fundamental questions regarding the use of neurostimulation to control neuronal networks in this temporally precise sense. That is, rather than simply stimulating the brain, the goal of this research is to develop new engineering theory and methods to allow practitioners to steer the activity in neural circuits so as to create complex patterns of activity, or neural spiking. Thus, this highly transdisciplinary project will elucidate enabling theory for the use of neurostimulation and will lead to new and fundamental contributions to systems theory and control engineering. The project will also support new initiatives to promote interdisciplinary education for students from traditionally underserved populations through the creation of summer workshops for students from local high schools in the city of St. Louis, MO.

By bridging ensemble systems theory with computational neuroscience, general and versatile frameworks for neuronal control will be formulated. Specifically, oscillator and conductance-based neural models will be used to mathematically model both oscillatory and non-oscillatory regimes in brain networks. Using these models, the proposed work will determine fundamental limits on the controllability of neuronal spiking or synchronization by the application of external inputs. The notion of ensemble reachability is also proposed and will be examined via entropic gain analysis and dynamic optimization. These characterizations of fundamental control properties in both oscillatory and non-oscillatory neural dynamic regimes will then facilitate the development of control design paradigms to synthesize optimal controls for the creation of complex patterns in neuronal populations, such as firing or entrainment patterns. Methods of ensemble control and formal averaging will be employed to derive optimal sequence and pattern controls, and stochastic versions of these problems will also be treated using stochastic control techniques to ensure tolerance to noise and uncertainty, which are pervasive in neural circuits. Thus, the results of the proposed research will include a unified systems-theoretic framework for analyzing the control of physiologically relevant brain networks; and further, will include a set of formal neural control design methods that may be readily translated to a range of neurostimulation implementations.

The study of neural coding asks how the brain converts raw signals into usable information that allows us to see, hear and think. Studying the mechanisms of neural coding is a persistent scientific challenge that has important implications for uncovering the workings of the brain. This award supports fundamental research that will enable a new approach to studying neural coding from the perspective of engineering theory. This perspective recognizes that networks in the brain can be modeled in terms of their physics and, thus, studied using many of the same tools that are used to study complex engineered systems such as aircraft and power grids. However, the brain possesses a level of complexity that far exceeds those of typical engineered systems and, consequently, existing engineering approaches must be adapted and augmented to meet biological realities. By addressing these gaps, this research will lead to new methods in engineering, advances in neural technology, and new ways of studying human brain function. This research is highly multi-disciplinary, involving systems engineering, mathematics and neuroscience. As part of the award, several new initiatives will be pursued to facilitate dialogue across these disciplines and to foster increased participation of underrepresented groups in engineering and science through the establishment of summer research internships for local high school students.

This award approaches neuronal networks through the lens of dynamical systems and control theory. This approach is based on the premise that understanding the input-output relationships of neuronal networks, mediated by their dynamics, will shed new light on fundamental questions in neuroscience, including the link between neural dynamics and information processing. In pursuit of this goal, the award focusses on two main objectives: First, neuroscientifically-motivated adaptions of systems theoretic properties, such as reachability, will be formulated so as to understand how dynamics govern neural input-output relationships. Since the connections in brain networks constantly adapt, emphasis will be placed on the notion of a brain state, which characterizes both the activity and the network structure at a given time. Second, control methods will be developed for the modulation of such states. To do so, a new class of objective functions will be defined in terms of the systems-theoretic properties conferred by the network. For example, such objectives will involve using controls to expand a network's reachable space, rather than just controlling its activity. The control input in this context may be quite general, and several specific scenarios, including neurostimulation, will be studied.

Project Summary Coma is a state of unconsciousness due to severe brain injury, in which patients are rendered unresponsive to external stimuli. Due to the limitations of current clinical tests in identifying a specific injury or causes associated with coma, devising treatment strategies for coma patients is a persistent clinical challenge. A signature feature of coma is severe disruption of the brain's electrical activity. Thus, the electroencephalogram (EEG), which measures the brain's electrical activity patterns, is routinely used in the neurology and neurosurgery intensive care unit (NNICU) to monitor patients in coma. However, the utility of EEG for diagnosing coma is largely limited to clinicians reading electrical activity in `raw' form as waveform tracings on a monitor. The primary goal of the proposed research is to develop and evaluate new algorithms, derived from engineering theory that will extract information about coma from the EEG that might not be apparent when reading the activity with the naked eye. Consequently, these new methods will enable the automatic EEG-based classification of coma etiology, gradation of injury severity, and prediction of clinical outcome. Eventually, these techniques could potentially be used to help tailor clinical treatment strategies for patients in coma. In this project, we will record EEG data from patients diagnosed with a range of coma etiologies. These data will be assimilated into a biological mathematical model for how the brain produces electrical activity, i.e., the neural dynamics. Enabled by these models, we will use a new type of analysis, called network reachability analysis, which characterizes the different types of electrical activity patterns that the models can produce. As an analogy, an airplane in flight might seem relatively stationary, but the plane's dynamics are actually complex since it could execute many different maneuvers at any time. Our analysis will describe how many `maneuvers' the brain is capable of making, thus providing a dynamical, quantitative characterization of the brain's lability. Our hypothesis is that different types of coma will exhibit different lability. To test this hypothesis, and to explore its clinical utility, we will apply network reachability analysis to the recordings we will obtain from patients with coma. Through this analysis, we will construct quantitative biomarkers that could be integrated into a new type of EEG monitor tailored for coma and other related disorders. Thus, the outcomes of this project will have significant and immediate impact on neurocritical care by facilitating more precise quantitative analysis of the neural dynamics of coma. More generally, the development of these techniques might shed new light on the mechanisms that underlie pathological states of unconsciousness, as well as normal sleep and wakefulness.

Project Summary The long terms goal of this project is to enable the control of large networks in the brain using neurostimulation technologies, a key focus of the BRAIN initiative. These technologies, including optogenetics, are developing at unprecedented rates and, consequently, are allowing scientists to make increasingly specific extrinsic perturbations to the activity in neural circuits. However, the nature of these perturbations remains largely limited so that the stimulated neuronal population is activated or deactivated en masse. As scientists seek to uncover the finer mechanisms of brain function, methods will be needed that allow more complex spatiotemporal activity patterns ? neural trajectories ? to be induced in these networks. The immense scale and interconnectedness of networks in the brain make this problem highly nontrivial. One may liken this problem to a musician on stage attempting to elicit a specific, unique response from each member of their audience individually, while playing to the group as a whole. To better understand these challenges and attempt to surpass them, our proposal introduces early concepts at the intersection of neuroscience and control theory, the mathematical study of how to optimally ?steer? complex systems subject to their dynamics, possible constraints, and an objective function that measures differences between the desired and induced trajectories. Our specific research aims are grounded in our team's interdisciplinary experience at the interface of dynamical systems, control theory and neuroscience. In Aim 1, we will study how the architecture and dynamics of networks in the brain enable control with respect to natural inputs, i.e., excitation through sensory pathways. In other words, we seek insights into how brain networks control themselves, towards better designing extrinsic stimulation. In Aim 2, we will develop a new toolkit, adapted from modern optimal control engineering, for designing neurostimulation input waveforms that are capable of creating high-dimensional trajectories (e.g., patterns of spikes) in large neuronal networks. In support of Aims 1 and 2, we will develop an innovative benchmark model containing structural and dynamical features pervasive in many salient neuronal networks. Finally, in Aim 3, we will perform in vivo experiments in which we will deploy our theoretical innovations to induce high-dimensional neuronal trajectories in a mouse somatosensory network using optogenetics. The proposed research will yield tangible outcomes in the form of new neurostimulation design methodologies and a benchmark control model that will be disseminated to the broader neuroscience community. Further, our theoretical developments are an important complement to continued growth in stimulation technology and cellular manipulation methods, facilitating a more complete approach to uncovering the mechanisms of the human brain.

This Faculty Early Career Award (CAREER) project is aimed at understanding cognitive functions in the human brain by investigating neural mechanisms of cognition using tools from systems and control theory. The research will develop mathematical models that will help understand internal brain dynamics. The study will focus on establishing connections between properties of neural circuits and their information processing functionality. The research will use brain activity data from human subjects to validate the newly developed mathematical models to establish connections between brain dynamics and cognitive functions. The outcomes of this project can provide insights into developing treatments for neurological conditions associated with altered cognitive functions such as loss of cognitive ability due to severe brain injuries. The project will also help to educate scientists working in the cross-cutting discipline of systems theory and neuroscience and will improve interpretability and accessibility of neurological diagnostics. The PI has well integrated educational outreach program for students from a St. Louis magnet school that includes use of neurotechnology kits for exploring and visualizing brain dynamics. This research will foster sustained excitement for scientific inquiry in mathematics and brain science.

This CAREER project will develop tools and techniques to understand certain cognitive states using methods of dynamics and control theory to analyze the dynamic behavior of neurons and the networks they form. The research will shed light on the role of dynamics and control within micro-scale neuronal networks as they relate to information processing. It will help understand how the dynamics of brain networks enable cognitively salient operations such as multi-modal integration and information retention. Along with these control-theoretic developments, this work will explore new methods for model- and data-driven characterization of neuronal dynamics at multiple spatial scales. The three specific objectives of this work are: (i) analyzing the relationship between macro-scale brain dynamics and cognitive function through new developments in dynamics and control; (ii) studying how neural dynamics may emerge in ways that support information processing; (iii) ensuring the long-term success of the program through numerous pedagogical and curricular endeavors.

This award supports basic research regarding the question of how networks in the brain allow odors to be detected and perceived. Such a question is of fundamental interest in neuroscience because responding to odors or scents is one of the most basic ecological abilities exhibited across different animal species. Further, responses to odors are highly dependent on context. For example, certain smells may create both attractive and repulsive reactions, depending on small differences in dilution or whether they are encountered alone or as components in a cocktail. Thus, studying how the brain processes odors can provide important clues regarding how animals and humans sense and perceive in complex environments. In seeking such understanding, this project uses a unique combination of methods from neuroscience, mathematics, and engineering. Brain activity from two different animal species are recorded during experiments in which odors are presented in isolation and in mixtures. Subsequently, data analysis and mathematical modeling is used to identify brain activity patterns that distinguish the reaction of the animals to the odors in question. Hence, the project uncovers how particular brain networks transform and transmit odor information in a way that is central to the sense of smell. To broaden the impact of these studies, the project includes the development of a summer internship in sensory neural engineering, intended to allow undergraduate and high school students to learn about and experience how different academic disciplines contribute to future brain science.

The extent to which sensory networks amplify or suppress perceived differences in odor valence remains a fundamental, unanswered question in sensory neuroscience. The overarching hypothesis of this project is that indeed, there exists a well-defined set of transformations, governed by neuronal dynamics, which map sensory network activity to behavior. Specifically, the project will determine: (a) How neural networks enable the formation of time-varying neural activation patterns, or, trajectories, in response to sensory stimuli, (b) The mapping from trajectories to behavioral outcome, and (c) The commonality of this mapping across species. The research goals use an interdisciplinary approach combining sensory systems neuroscience in two species, locusts (Schistocerca americana) and round worms (C. elegans), with computational modeling and dynamical systems theory. Neural and behavioral responses are recorded from animals receiving nominally attractive and aversive odors, and these data inform computational models of the sensory networks and ensuing behaviors. The models generate predictions on how behavioral responses might be modulated by a change in selectivity, or background state. The latter is tested through a paradigm wherein animals are systematically fed or starved, thus shifting their response dynamics on the aversive-attractive spectrum. Subsequently, model-based sensitivity analyses is used to predict mixture response curves and paradoxical mixtures (e.g., two aversive stimuli that when mixed, elicit an attractive response). These predictions are tested by delivering component stimuli in systematic ratios. Thus, the overall methodology combines physiological experiments with new systems-level analysis in an integrated, multidisciplinary modeling-theory loop.

The goal of this workshop is to bring together researchers from the dynamics & control and neuroscience communities with the purpose of identifying promising future research directions focusing on the use of dynamical systems and control theory to study the workings of the human brain. The workshop will be held in Washington University in St. Louis on June 25-27, 2017.

Recent years have witnessed substantial cross-cutting interest in the use of dynamical systems and control theory in the study and engineering of biological systems. The goal of this workshop is to provide a focused forum for the discussion of research synergies between experts from the dynamics, control and neuroscience communities. By facilitating this discussion early in the inception of this interdisciplinary area, the workshop will aid in ensuring long-term, sustained research impact. The workshop will broaden the reach of the systems and control community by creating new opportunities for synergy with the neuroscience community. The workshop will also emphasize student participation through a poster session and involvement in organizing the workshop through a student organizing committee.

NSF 1835209
NCS-FO: Modeling Individual Differences in Cognitive Control as Variation in Neural Activation Trajectories
Abstract:
This award supports fundamental research to examine how activity within brain networks allows humans to adapt their behavior in order to achieve goals and complete mental tasks. Such processes within the brain, referred to as cognitive control, are thought to differentiate individuals in terms of mental abilities that are critical for successful navigation in activities of daily life, such as planning, problem solving and reasoning. Current brain imaging methods enable examination of the activity and interactions among brain networks as individuals perform various tasks, thus providing a window into the mechanisms of cognitive control. However, research efforts to date have mostly used imaging data to generate snapshots of brain activity that are averaged across groups of individuals and many different events while performing a task. In this research program, the investigators develop a new form of analysis to characterize the moment-to-moment fluctuations in brain activity, within each individual, as they transition from rest to cognitively demanding task conditions. In particular, efforts will be directed towards the development of a computational model that can predict how brain networks coordinate activity over seconds-level timescales in response to changing task conditions. A key aspect of the effort will be to develop unique models for each individual, drawing from a large database of previously obtained neuroimaging data. In these data, individuals perform a range of tasks requiring different cognitive control strategies, some proactive (sustained) versus others reactive (transient). Thus, application of the model will reveal how the brains of these individuals differentially respond to various types of cognitive demand. The development of this model also provides a unique opportunity for education and outreach; specific efforts will be directed toward the development of a software platform through which members of the public can work with demonstration models to probe and learn about how different patterns of brain activity relate to cognitive function.

Functional neuroimaging has allowed for detailed spatial and temporal characterizations of brain network activation in an effort to elucidate the neural underpinnings of cognitive control. However, such analyses usually rely on static snapshots of neural activation patterns in individual brain regions and/or correlational indices of inter-regional co-activation (i.e., functional connectivity). Further progress in understanding distinctions between cognitive states and cognitive control strategies requires more precise descriptions of the brain dynamics that govern how patterns of neural activity (trajectories) evolve across time. Leveraging recent advancements in optimization theory that allow for reliable high-dimensional parameter estimation, this award will support the validation and parameterization of single-subject dynamical models using high-resolution, long-duration resting-state fMRI data from the Human Connectome Project, which contains data from over 1000 individuals. Subsequent model analysis will characterize individual differences in terms of brain network dynamics, focusing on quantitative metrics of the ruggedness of the attractor landscape (which indicates the diversity of achievable trajectories) and the consequent energetic costs incurred by shifting between cognitive states and strategies. Hypothesis testing will be conducted with a unique follow-up dataset, consisting of a subset of HCP participants and monozygotic (identical) twins (over 100 in total) tracked in multiple neuroimaging sessions, under conditions that systematically manipulate cognitive control strategies.

This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

ABSTRACT Short-term working memory is critical for all cognition. It is important to fluid intelligence by definition and is disordered in many psychiatric conditions. It is also an ideal model system for studying the link between the dynamics and functions of neural circuits. Short-term storage requires dynamics that are flexible enough to allow continuous incorporation of new information, yet stable enough to retain information for tens of seconds. Much is known about the neuronal substrate of short-term memory. There is a gap, however, in our knowledge of how neuronal resources are efficiently allocated to store multiple items. This gap is particularly striking given that a multi-item memory task (memory span task) is often used to measure fluid intelligence. Neurons in frontal areas are active during a memory period, and individual neurons are tuned to respond to particular memoranda. It is known that individual cells ramp up or down during a memory period. However, we were surprised to discover in preliminary experiments that 80% of individual cells in memory circuits lose their tuning before the end of a 15s memory period. This loss of tuning occurs at similar times across repeated trials; a neuron that loses tuning at 3s in one trial seldom remains tuned for more than 7s in a subsequent trial, and vice versa. This leads to the question of whether cells with common ?drop-out? times are linked together in a subnetwork, similar to the ?slot? organization often posited to support multi-item memory. We formulated a theory about how these subnetworks might be organized to enact a form of efficient resource allocation that balances demand for memory capacity against memory duration. The primary goal of this proposal is to test the validity of this theory, and more generally probe memory circuits for evidence of functional subnetworks, using a unique combination of long-delay multi-item memory tasks, computational modeling and analysis. In Aim 1, we will test a key facet of our theory: how storage of information may interact with the phenomenon of ?drop-out? to either help or hinder short-term memory storage. In Aims 2 and 3, we will test whether cells with similar dropout times also share other properties, as a way of determining whether they are in fact linked together in subnetworks. In parallel, we will develop modeling and optimization tools to ask how such subnetworks might be enacted in neural circuits, while also engaging the higher-level question of whether subnetworks are in fact a sensible solution to the problem of efficient resource allocation in the first place. A key aspect of the proposal is the integration of experimental and computational methods, including formalisms from information and control theories, so as to build tight links between (i) the observed phenomenology; (ii) the mathematical consistency of the theory; and (iii) how (i) and (ii) might be reconciled mechanistically in the dynamics of neural circuits. Together, these Aims have the potential to change the way we think about the neuronal substrate of short-term memory and how neural circuits are structured to best manage their resources.