Steven J. Schiff - US grants

The well-known extreme variability of spinal cord reflex output has defied analysis suing classical statistics. This variability renders the interpretation of electrophysiologic monitoring during spasticity surgery extremely difficult. Preliminary studies have re-evaluated reflex output data using new techniques developed for the analysis of complex phenomena. In the normal human, it appears that reflexes fluctuate on many different time scales: seconds, minutes, and perhaps hours. Physical systems which fluctuate on many time scales are "self-similar", and this is the hallmark of a "fractal" process. The results generate the following working hypothesis: time-series of spinal cord reflex output fluctuate with a fractal pattern. this idea will be rigorously tested using data collected from the spinal cord of the decerebrate cat including: 1) neuronal population response, 2) individual neuron firing frequency and 3) the tendon force generated. Data will be obtained while varying the frequency of stimulation and with the reflex feedback loop opened and closed. to characterize fractal behavior, non-linear analytical methods will be extensively used. By employing simulated data sets as mathematical controls, these methods should help determine whether the observed fractal patterns originate from deterministic or stochastic processes. The results of this work are important from both clinical and basic science perspectives. Clinically, proving that spinal cord reflexes fluctuate on time scales far longer than the measuring period permitted in the operating room would lead to a radical shortening of operative time and risk. On a basic level, the results will provide information leading to a deeper understanding of the origin of apparently random fluctuations in a simple input-output neural circuit in the mammalian nervous system. It is expected that the results of this study will be applicable to more complex neural circuits in the central nervous system.

IBN 97-27739 SO, SCHIFF, GLUCKMANN. An understanding of synchronous activities within an ensemble of neurons is essential in the study of neuroscience. It is important to understand and characterize both the computation within an ensemble, as well as the information flow between different ensembles within the brain. In the so called "binding problem", when spatially disparate neurons must coordinate to compute aspects of sensory perception, synchrony is essential. Traditionally, these issues have been addressed using the concept of identical synchrony (IS) which assumes that two or more ensembles of the brain are performing the same activities in locked time step with each other. However, in ensembles with generic nonlinear components, of which neuronal ensembles are most certainly included, more complex coherent behaviors should arise. Consequently, our concept of dynamical coherence beyond identical synchrony must be broadened. Chaos theory broadly encompasses the study of such nonlinear dynamical systems. A major theoretical advance in this field was the recognition that seeming erratic behaviors from these nonlinear systems could be effectively characterized by a set of special unstable equilibrium states. In a cartoonist view, these so called unstable periodic orbits (UPOs) are hills and valleys of an abstract dynamical landscape. As the system progresses in time, the state of the systems can be described by a trajectory within this dynamical landscape constructed with the UPOs. For coupled systems (neurons), the arrangement and symmetry of these hills and valleys reflect the varying degree of dynamical coherence exhibited within the system. Most importantly, analogous to statistical mechanics in physics, these UPOs form a framework of microscopic states for the system and their structural changes afford a description for the topographical changes within this dynamical landscape. A thermodynamical description based on these UPOs for the various possible dynamical coherent states might then be constructed. Theoretical tools developed will be applied to quintessential examples of neuronal coupling from our archived biological data: two coupled neurons and two ensembles of neurons. Results from this project will both theoretically broaden our understanding of coupled nonlinear oscillators, including neurons, coupled mechanical and electronic devices, etc., and will serve as the initial attempt to experimentally characterize the grammatical code used between ensembles of neurons.

DESCRIPTION (provided by applicant): The physics of pattern formation has made dramatic strides over the past 20 years. Nevertheless, the ability to apply this knowledge to neuronal systems has been limited by our knowledge of the dynamics of neuronal interactions, and our lack of a system parameter to control such patterns. Over the previous period of this R01, we have demonstrated that electric fields can serve as a feedback parameter to adaptively control patterns of neuronal activity. Furthermore, we have demonstrated that the interactions between individual neurons as they form patterns can be characterized in detail. This renewal proposal will seek to test the Hypothesis that Parametric control of neuronal patterns can be achieved with electric field feedback. Such feedback can serve as a basic tool to explore how neuronal ensembles dynamically form and change their patterns of activity, and serve as a means to interact with and selectively modify pathological neuronal dynamics. The Specific Aims for this research plan are to develop intelligent interaction and control strategies for neuronal activity patterns using 1) In Vitro experiments to determine how the intracellular interactions between neurons determine the nature of the transition between patterns under the influence of feedback electric fields, 2) develop a compartmental computational model to directly simulate these experiments, 3) to apply insights gained from the compartmental model to design and test rational control strategies for seizures and theta rhythm in In Vitro and In Vivo experiments. The long term goals are 1) basic science: to understand neuronal pattern formation and control from a physical point of view. I hope to exploit this insight to create more rational modulation and control strategies to probe the dynamics of neuronal networks, and 2) medical: to lay a foundation for interfacing with and controlling pathological brain dynamics without requiring surgical destruction of parts of the brain.

DESCRIPTION (provided by applicant): This CRCNS application derives from work performed in a current DAAD (Deutscher Akademischer Austausch Dienst, German Academic Exchange Service) Grant between the Technical University of Berlin and Penn State University entitled: "Feedback control of spreading depolarizations in neural systems: Theory and Experiments". The design of this CRCNS proposal, and all preliminary data, were generated during the course of German Faculty and PhD students coming to Penn State University, and the synergistic collaborative efforts to establish the feasibility of feedback control of spreading depression. Spreading depression (SD) is a dramatic depolarization of brain that propagates slowly and is the physiological underpinning of the initial aura in migraines. The following hypothesis is posed: SD can be represented in computational models of the underlying neuronal biophysics, and can therefore be controlled using model-based control strategies. The project starts by developing an experimental preparation using a tangential 2-dimensional visual cortex rodent brain slice. SD is triggered with a perfusate potassium perturbation, and SD is imaged using a sensitive CCD camera that detects the intrinsic optical imaging signal associated with index of refraction changes from cellular swelling. A model-based strategy similar to that used in autonomous robotics such as airframe autolanders is employed. A hardware and software control system takes the optical image in real-time, fuses it with a model of SD, reconstructs the underlying physiological processes, calculates needed control, and modulates an electrical field to modulate SD. Both biophysically accurate models of the neuronal compartments and ion flows, and reduced models that reflect the dynamics of the wave propagation, will be used as observation and control models. Intellectual merit: This will be the first experimental demonstration of model-based control of a neuronal network. Similar engineering strategies have revolutionized advanced robotics, and the fundamentals learned from a fusion of computational neuroscience with control engineering will have wide ranging adaptations in other areas of neuronal modulation. Furthermore, this will be the first model-based control of a physiological mechanism that underlies a dynamical disease of the brain - migraine auras. The control models will further serve as probes to gain increased understanding of the mechanisms of SD. The team assembled has a substantial track record in the range of disciplines required to carry out this project: neurophysiology, experimental and theoretical physics, computational neuroscience, control theory and neural engineering. The preliminary work shown in the proposal suggests that this project is feasible given the resources requested. Broader impact: Fusing computational neuroscience models with modern model-based control theory will lay the foundation for a transformational paradigm for the observation of activity within the brain, as well as access to a more optimal technology for the control of pathological processes in the brain. A transdisciplinary German-American educational collaboration will be formed where the graduate students trained (and the PIs) will synergistically work together within the interface between computational neuroscience, control theory, experimental neurophysiology, and control system engineering. The PIs have a track record in training and mentoring women and underrepresented minorities, and they will make every effort to seek such trainees for the mentoring opportunities of this project. As a collaborative partnership, the PIs anticipate that what is learned in controlling SD may provide a set of testable strategies for electrical control of migraines in people who suffer from severe migraine attacks and are pharmacologically intractable. Furthermore, based upon this CRCNS, the same science and engineering will be applicable to the modulation of oscillatory waves and rhythms in both in vitro (e.g. Schiff et al 2007) and in vivo (e.g. Sunderam et al 2009) systems. They plan to widely disseminate the algorithms and hardware design developed as described in the Data Management Plan.

DESCRIPTION (provided by applicant): Sleep is a fundamental biological cycle that is coupled into every aspect of body function from behavior and information processing to metabolic storage and release. Sleep-wake patterns correlate with, and sleep disruptions are comorbid with, many neurological and mental health disease dynamics including epilepsy. Abnormal sleep can be disruptive to quality of life and further exacerbate the primary disorders. Within the past decade a number of groups have developed mathematical and computational dynamical models for the network of brain nuclei and cell groups that regulate sleep-wake dynamics. But their validation to date has been substantially limited to reproduction of statistic of cycle time and dwell time durations, and their application to understanding and control of diseases limited. The first objective of this project is to validate and optimize these models for reconstruction, forecasting, and control of sleep-wake regulation. This involves experimentally recording activity from select cell groups of the sleep-wake regulatory system (SWRS) along with cortical, hippocampal, and behavioral activity. The mathematical models will be incorporated into model-based data assimilation (DA). The parameters and models will be optimized for reconstruction and forecasting, and performance will be used to establish the 'best' model. Experimental perturbation of sleep state and sleep cycle dynamics will be done with both sensory and direct neural stimulation. The models will then be modified to account for and predict changed dynamics from such perturbations. The second objective of this project is to apply these models and framework to understand and control sleep-cycle dis-regulation in a model of temporal lobe epilepsy. This involves experimentally recording activity from the SWRS in epileptic animals, modifying and optimizing the models to reconstruct and forecast the observed sleep cycle dynamics. The models will then be used in closed feedback form to prescribe control perturbations to regularize the sleep cycles of the epileptic animals. The project embodies a paradigm shift for neuroscience and neural-engineering in which computational models are validated and optimized through their capacity to reconstruct and forecast detailed time series from real neurological measurements, that such model-based reconstruction is used to observe detailed state dynamics from less costly (invasive or damaging) measurements, and in which such biologically based models are used to control neurological systems and treat neurological disorders. The approach of this proposed research will have a major impact in diagnosing, monitoring, and controlling neurological disorders by both incorporating detailed biologically based models into the measurement or observation process, and by allowing remote observation through measurement of identified less costly measurements. The specific validation and improvement of computational models and observation methodology of the sleep-wake regulatory system will allow detailed investigation of its role in a host of neurological diseases in which sleep regulatin is implicated either as a cause or consequence, such as epilepsy and schizophrenia, and thereby the development of interventions or therapy. In addition to the theoretical and experimental advances, educational and outreach will be served through this project, including development of new course materials and enhancing underrepresented participation in research.

DESCRIPTION (provided by applicant): I propose a novel approach to institute model-based feedback control to seek a rational, and optimal, framework to reduce neonatal sepsis (NS) in developing countries. In addition to reducing deaths from NS, one of the leading global killers of children worldwide, this will reduce the sequelae in the survivors of NS such as postinfectious hydrocephalus (PIH) - likely the dominant cause of hydrocephalus in children worldwide. I presently am a PI on a Phase III NIH sponsored randomized controlled surgical trial in Africa seeking to optimize treatment of PIH (clinical trial # NCT01936272). However, as an alternative to surgical treatment of children with irreparable brain damage, I am now in a unique position to learn how to more effectively treat NS in this setting, and thereby better prevent the large numbers of infants with PIH in this part of the world that we would otherwise need to surgically palliate. Although a pediatric neurosurgeon, I have acquired considerable expertise in both control engineering and physics. I recently wrote the first book on Neural Control Engineering, published by the MIT press in 2012. This Pioneer Award seeks to leverage my knowledge of control engineering and apply this to an entirely new avenue of research for me - seeking to impact in a sustainable way both the morbidity and mortality of NS. I have put in place a unique infrastructure in Uganda to enable this project. I have secured Ugandan medical licensure. I have organized two pilot projects with two key institutions: a pediatric neurosurgery specialty hospital in Mbale (the CURE Children's Hospital of Uganda), and a major referral hospital in Mbarara (at the Mbarara University of Science and Technology). At Mbale, most of the hydrocephalus in Uganda is now treated, and the majority of these cases are postinfectious following NS. At Mbarara, the most common admission to their infant ward is NS, and all of their hydrocephalus is referred to Mbale. At each i

Challenge, Innovation and Impact: In recent years, we have demonstrated that it is feasible to predict epidemic disease outbreaks from retrospective seasonal and geographical case data and to show that we can take climate factors into account in our predictive models. We are moving closer to real-time prediction at the population level. But we have never used prediction at point-of-care for treating the individual patient. Presently, personalized medicine uses delayed results of laboratory testing of individuals. For infectious disease, most of such testing has targeted the pathogen in the host-pathogen interaction. The role of laboratory testing is to modify therapy after a variable period of time delay. Personalized medicine today is reactive. Complicating matters further, many infectious epidemic diseases are strongly dependent on environmental factors and climate. Lastly, we want to name the pathogens we are fighting, but we really need to know the resistance characteristics to select therapy for patients effectively. Both speciation and resistance can now be determined from molecular data, which can be integrated into point-of-care treatment predictions. We here propose a radically different approach to the treatment of infectious diseases. Our hypothesis is that the alternative to time-delayed and expensive laboratory analysis of specimens from individual patients, is to use predictive modeling to forecast point-of-care treatment. Time-delayed personalized testing can be conducted as surveillance, and that data used for real-time prediction to guide point-of-care treatment. We will introduce predictive personalized public health (P3H) policy at the individual patient level, with the potential to substantially improve patient outcomes compared with our present reactive approaches. Our key rationale is to expand population infectious disease predictive modeling in order to achieve prediction for treatment at point-of-care. Our primary insight is that we can reposition the delayed reactive personalized testing from the urgent medical decision-making process, and into a predictive modeling framework. The gaps and opportunities in technology that we will address are four-fold. First, we will employ individual case geospatial mapping at a fine scale to take into account infection spread and environmental factors. Second, our ability to perform pan-microbial analysis using molecular techniques is now feasible. Third, modeling our novel fusion of data has no simple low-dimensional solution ? but machine learning technologies are now capable of handling such big data assimilation, model discovery and prediction. Fourth, our proposal is not an academic exercise. We have a partnership with the economic planners within a developing country to design and implement our new methods. We will prospectively tune and validate our algorithms in real-time. Our deliverable will be an open-source framework ready for clinical trials testing and adaptation to the public health infrastructure in any country.