Biohybrid Systems E-Book

• Herausgeber: Wiley-VCH
• Kategorie: Wissenschaft und neue Technologien
• Sprache: Englisch

Seitenzahl: 408

Veröffentlichungsjahr: 2012

Beschreibung

Accessible for engineers and physicists alike, this high-level work on this hot topic in medical physics covers the issue from principles of neuroscience to in-vivo use. Backed by extensive data tables and software samples on the Web.

Leseprobe

Contents

Cover

Related Titles

Title Page

Copyright

Preface

List of Contributors

Chapter 1: Merging Technology with Biology

1.1 Introduction

1.2 NeuroDesign

1.3 The NeuroDesign Approach

1.4 Neuromorphic Control of a Powered Orthosis for Crutch-Free Walking

1.5 Frontiers of Biohybrid Systems

1.6 Chapter Organization

Acknowledgment

References

Chapter 2: Principles of Computational Neuroscience

2.1 Introduction

2.2 Some Physiology of Neurons

2.3 General Formalisms in Neuronal Modeling

2.4 Synaptic Coupling and Plasticity

2.5 Computational Models of Neuronal Systems for Biohybrid Applications

2.6 Resources

Acknowledgments

References

Chapter 3: Neuromorphic Electronic Design

3.1 Choices for Neuromorphic Circuits: Digital versus Analog

3.2 The Breadth of Neuromorphic Systems

3.3 The Fundamental Processing Unit: The Neuron

3.4 Sensing the Environment

3.5 Conclusions

3.6 Resources

References

Chapter 4: Principles of Neural Signal Processing

4.1 Introduction

4.2 Point Process Theory

4.3 Analyzing a Point Process

4.4 Dynamic Neural Processing

4.5 Information Theory and Neural Signal Processing

4.6 Summary

References

Chapter 5: Dynamic Clamp in Biomimetic and Biohybrid Living-Hardware Systems

5.1 What is a Dynamic Clamp?

5.2 Dynamic Clamp Performance and Limitations

5.3 Experimental Applications of Dynamic Clamp

5.4 Dynamic Clamp System Implementations and Future

5.5 Resources

References

Chapter 6: Biohybrid Circuits: Nanotransducers Linking Cells and Neural Electrodes

6.1 Introduction to Neural–Electrical Interfaces

6.2 Neural Probes with Nanowires

6.3 Microelectrode Arrays with Carbon Nanofibers

6.4 Microelectrode Arrays with Carbon Nanotubes

6.5 Microelectrode Arrays with Conducting Polymer Nanomaterials

6.6 Nanoelectrodes for Neural Probes

6.7 Summary and Future Work

Acknowledgment

References

Chapter 7: Hybrid Systems Analysis: Real-Time Systems for Design and Prototyping of Neural Interfaces and Prostheses

7.1 Introduction

7.2 Technology

7.3 Applications

7.4 Hybrid Systems Analysis in the Leech Heart Interneuron

7.5 Discussion

Acknowledgments

References

Chapter 8: Biomimetic Adaptive Control Algorithms

8.1 Introduction

8.2 Biomimetic Algorithms

8.3 Discussion

8.4 Future Developments

References

Chapter 9: Neuromorphic Hardware for Control

9.1 Neuromorphic Hardware for Locomotion

9.2 Neuromorphic Hardware for Audition

9.3 Neuromorphic Hardware for Vision

9.4 Conclusions

References

Chapter 10: Biohybrid Systems for Neurocardiology

10.1 Introduction

10.2 Autonomic Neural Control of the Heart

10.3 Monitoring and Modulating the Autonomic Reflexive Control of the Heart

10.4 Conclusions

References

Chapter 11: Bioelectronic Sensing of Insulin Demand

11.1 Sensor Technologies and Cell Therapy in Diabetes: A Life-Long Debilitating Disease

11.2 The Biological Sensor: Function of β-Cells and Islet

11.3 Automated Islet Screening and Bioelectronic Sensor of Insulin Demand

11.4 Closed Loop Exploration In Vitro

11.5 Methods

11.6 Results

11.7 Conclusions

References

Index

Related Titles

Wandelt, K. (ed.)Surface and Interface Sciences Volume 1: Basic Concepts and Methods approx. 500 pages Hardcover ISBN: 978-3-527-40488-9

Schuster, H. G. (ed.)Reviews of Nonlinear Dynamics and Complexity Volume 3 260 pages with approx. 59 figures 2010 Hardcover ISBN: 978-3-527-40945-7

Akay, M. (ed.)Handbook of Neural Engineering 662 pages 2007 Hardcover ISBN: 978-0-470-05669-1

Kumar, C. S. S. R. (ed.)Nanodevices for the Life Sciences 489 pages with 211 figures and 10 tables 2006 Hardcover ISBN: 978-3-527-31384-6

Waser, R. (ed.)Nanoelectronics and Information Technology 995 pages with 1148 figures and 47 tables 2005 Hardcover ISBN: 978-3-527-40542-8

All books published by Wiley-VCH are carefully produced. Nevertheless, authors, editors, and publisher do not warrant the information contained in these books, including this book, to be free of errors. Readers are advised to keep in mind that statements, data, illustrations, procedural details or other items may inadvertently be inaccurate.

Library of Congress Card No.: applied for

British Library Cataloguing-in-Publication Data

A catalogue record for this book is available from the British Library.

Bibliographic information published by the Deutsche Nationalbibliothek

The Deutsche Nationalbibliothek lists this publication in the Deutsche Nationalbibliografie; detailed bibliographic data are available on the Internet at http://dnb.d-nb.de.

© 2011 WILEY-VCH Verlag & Co. KGaA,

Boschstr. 12, 69469 Weinheim, Germany

All rights reserved (including those of translation into other languages). No part of this book may be reproduced in any form – by photoprinting, microfilm, or any other means – nor transmitted or translated into a machine language without written permission from the publishers. Registered names, trademarks, etc. used in this book, even when not specifically marked as such, are not to be considered unprotected by law.

Cover Design Adam-Design, Weinheim

Print ISBN: 978-3-527-40949-5

ePDF ISBN: 978-3-527-63938-0

oBook ISBN: 978-3-527-63936-6

ePub ISBN: 978-3-527-63937-3

Preface

In the last decade of the 20th century, the “Decade of the Brain”, the scientific community put forth a concerted effort towards understanding the nervous system. Although experimental neurophysiological approaches provided many advances, it became increasingly evident that mathematical and computational techniques would be required to achieve a comprehensive and quantitative understanding of neural system function. “Computational Neuroscience” emerged to complement experimental neurophysiology. Simultaneously, fueled by engineering breakthroughs, the last two decades have seen a phenomenal rise in our ability to probe the nervous system and to influence neural system activity across scales of complexity and states of disease. Devices that use focused electrical stimulation to activate neural circuits are now routinely used to restore hearing to the deaf and to alleviate the symptoms of Parkinson's disease, while emerging technologies will provide amputees with the ability to feel with their artificial limb. In the first decade of the 21st century, this new engineering paradigm that links living with non-living systems to investigate, intervene and harness neural plasticity to counter disease and disablement emerged in the form of “Neural Engineering”.

This book presents a window into the convergence of Computational Neuroscience and Neural Engineering. Over the past two decades it has been my privilege to be enriched by the flourishing of both Computational Neuroscience and Neural Engineering and to have the opportunity to dialogue with neuroscientists, mathematicians, physicists, and engineers from around the world. Two summers have played an important role in my personal engagement with these fields. One was a summer at Woods Hole, attending the ‘Methods in Computational Neuroscience Course’. Here, I listened to John Rinzel present phase space analyses methods, talked to Ron Calabrese about leech heart interneurons that I modeled, heard about the newly devised ‘Dynamic Clamp’ from Eve Marder, talked about ‘Consciousness’ with Christof Koch and others on the beach at night, and met a neuroscientist who became my postdoctoral mentor - Avis Cohen. It was Avis who suggested a summer at Telluride at the ‘Neuromorphic Engineering’ workshop. There, I listened to Rodney Douglas and Misha Mahowald, once again Christof Koch, and got introduced to the world of engineers trying to capture the biological neuron in hardware. It is not surprising then, that as a biomedical engineer fascinated by the two fields, I have sought to find a practical interface that is driven by the merger of the software and hardware models of neurons with the nervous system itself. It is at the summer courses that I met many of my fellow scientists and engineers who have over the years sought similar goals, some of who have contributed to this book.

Growth of such a transdisciplinary effort required a concerted investment by many institutions that were guided by people with foresight and boldness. Dennis Glanzman and Yuan Liu from the National Institutes of Health, USA and Kenneth Whang from the National Science Foundation, USA have played an unrelenting role in supporting programmatic growth of Computational Neuroscience and the research effort of several investigators. The Collaborative Research in Computational Neuroscience Program has supported a wide range of research efforts that underlie the development of biohybrid systems and has allowed me to seek new knowledge in spinal organization for motor control after spinal cord injury. The book and I have also benefitted from transdisciplinary dialogue on biohybrid systems and neuromorphic design at a series of workshops that we conducted with support through the Science of Learning Centers program at the National Science Foundation, USA under Soo-Siang Lim. Grace Peng from the National Institutes of Health has been a steady champion of programmatic growth in neural engineering and has been a supporter of the efforts of many, including me, in bringing technology to the people that stand to benefit from this technology. Most interestingly, Elmar Schmeisser from the Army Research Office saw promise in our work on neuromorphic control of spinal interfaces in the lamprey as the basis for a novel approach to control powered or thoses for people with lower limb dysfunction. It was a presentation of these multiple related areas of research that caught the attention of Wiley and I thank them for inviting me to develop a book to present our ideas about this emerging field of biohybrid systems. The growing interest in this topic motivated my colleagues and meto develop a book for a cross-section of scientists and engineers. We hope that this book will enhance the communication between computational neuroscientists and neural engineers and bring to attention the exciting new applications that biohybrid systems could offer clinicians who are eager to deliver new solutions to their clients. It has been my pleasure to have worked with the authors of the different chapters and their teams in the writing of the book. I thank them for their effort and for their enthusiasm, not only in penning their own chapters, but also in providing helpful critiques of others.

I must thank my brother Vikram who has over the many years shared with me many of his management skills that have allowed me to juggle multiple projects and work across academic-clinical-industrial partnerships. My parents, Sarla and Padam, are a steady source of support and guidance. My husband Jimmy and son Nikhar, who are both contributors to this book, have been my sounding boards, have withstood my immersion in various projects, but most importantly have been a never-ending source of joy and companionship. Finally, I am forever indebted to my doctoral thesis advisor, Peter Katona who fostered inquiry across boundaries, supported my inquisitiveness and nurtured my foray into new realms.

June 30, 2011 Miami, Florida

List of Contributors

James J. AbbasArizona State UniversitySchool of Biological and HealthSystems EngineeringCenter for Adaptive Neural SystemsP.O. Box 874404Tempe, AZ 85287USA

William BarnettGeorgia State UniversityNeuroscience Institute100 Piedmont Ave. SEAtlanta, GA 30302USA

Guilherme BontorinUniversité de Bordeaux 1Integration from Materialto System (IMS)IPB, UMR CNRS 5218351 Cours de la LibérationF-33405 TalenceFrance

Yannick BornatUniversité de Bordeaux 1Integration from Materialto System (IMS)IPB, UMR CNRS 5218351 Cours de la LibérationF-33405 TalenceFrance

Rik BuschmanMedtronic NeuromodulationHeerlenThe Netherlands

Linfeng ChenUniversity of ArkansasDepartment of Electrical EngineeringFayetteville, AR 72701USA

Richard CornelussenMedtronic NeuromodulationHeerlenThe Netherlands

Sharon CrookArizona State UniversitySchool of Biological and HealthSystems EngineeringCenter for Adaptive Neural SystemsTempe, AZ 85287USA

and

Arizona State UniversitySchool of Mathematical and Statistical Sciences and School of Life SciencesTempe, AZ 85287USA

Gennady CymbalyukGeorgia State UniversityNeuroscience Institute100 Piedmont Ave. SEAtlanta, GA 30302USA

Ralph Etienne-CummingsThe Johns Hopkins UniversityDepartment of Electrical andComputer Engineering105 Barton Hall, 3400 N. Charles StreetBaltimore, MD 21218USA

Robert E. HarbaughPennsylvania State HersheyMedical CenterCollege of MedicineDepartment of NeurosurgeryHershey, PA 17033USA

Ryan HooperGeorgia Tech & Emory UniversityWallace H. Coulter Department of Biomedical Engineering313 Ferst DriveAtlanta, GA 30332USA

Don H. JohnsonRice UniversityDepartment of Electrical andComputer Engineering6100 Main StreetHouston, TX 77005USA

Ranu JungFlorida International UniversityCollege of Engineering and ComputingDepartment of Biomedical Engineering10555 W. Flagler StreetMiami, FL 33174USA

and

Arizona State UniversitySchool of Biological and HealthSystems EngineeringCenter for Adaptive Neural SystemsTempe, AZ 85287USA

Lilian KornetMedtronic NeuromodulationHeerlenThe Netherlands

Jochen LangUniversité de Bordeaux 1Institut Européen de Chimie & Biologie (IECB)UMR CNRS 52482 rue Robert EscarpitF-33607 PessacFrance

Kevin MazurekThe Johns Hopkins UniversityDepartment of Electrical andComputer Engineering105 Barton Hall, 3400 N. Charles StreetBaltimore, MD 21218USA

Simone C.M.A. OrdelmanUniversity of TwenteMIRA Institute for Biomedical Technology and Technical MedicineFaculty of Electrical Engineering, Mathematics and Computer ScienceDepartment of Biomedical Signalsand Systems7500 AE EnschedeThe Netherlands

Astrid A. PrinzEmory UniversityDepartment of BiologyO. Wayne Rollins Research Center1510 Clifton Road NEAtlanta, GA 30322USA

Matthieu RaouxUniversité de Bordeaux 1Institut Européen de Chimie &Biologie (IECB)UMR CNRS 52482 rue Robert EscarpitF-33607 PessacFrance

Sylvie RenaudUniversité de Bordeaux 1Integration from Material to System (IMS)IPB, UMR CNRS 5218351 Cours de la LibérationF-33405 TalenceFrance

Malathi SrivatsanArkansas State UniversityArkansas Biosciences InstituteDepartment of Biological SciencesJonesboro, AR 72401USA

Francesco TenoreThe Johns Hopkins UniversityDepartment of Electrical and Computer Engineering105 Barton Hall, 3400 N. Charles StreetBaltimore, MD 21218USA

Vijay K. VaradanUniversity of ArkansasDepartment of Electrical EngineeringFayetteville, AR 72701USA

and

Pennsylvania State HersheyMedical CenterCollege of MedicineDepartment of NeurosurgeryHershey, PA 17033USA

Peter H. VeltinkUniversity of TwenteMIRA Institute for Biomedical Technology and Technical MedicineFaculty of Electrical Engineering, Mathematics and Computer ScienceDepartment of Biomedical Signals and Systems7500 AE EnschedeThe Netherlands

Sharmila VenugopalArizona State UniversitySchool of Biological and Health Systems EngineeringCenter for Adaptive Neural SystemsTempe, AZ 85287USA

R. Jacob VogelsteinJohns Hopkins University Applied Physics Laboratory11100 Johns Hopkins RoadLaurel, MD 20723-6099USA

Jining XieUniversity of ArkansasDepartment of Electrical EngineeringFayetteville, AR 72701USA

Hargsoon YoonUniversity of ArkansasDepartment of Electrical EngineeringFayetteville, AR 72701USA

Chapter 1

Merging Technology with Biology

Ranu Jung

1.1 Introduction

The most important trend in recent technological developments may be that technology is increasingly integrated with biological systems. Many of the critical advances that are emerging can be attributed to the interactions between the biological systems and the technology. The integration of technology with biology makes us more productive in the workplace, makes medical devices more effective, and makes our entertainment systems more engaging. Our lives change as biology and technology merge to form biohybrid systems.

This book describes some of the recent advances and some of the key challenges faced by engineers and scientists developing biohybrid systems that interface nerves, muscles, and machines. Modern computers have high computational capacity and high rates of internal information transfer between components; similarly, neurobiological systems have high computational capacity and high interconnectivity of neural structures. Some of the key developments in biohybrid systems have been in opening lines of communication between the engineered and the biological systems. Real-time communication between a nervous system and a device is now possible, but full and reliable integration is still far from reality. In order to achieve more complete integration, some of the key challenges in biohybrid system development are to improve the quality, quantity, and reliability of the information that can be transferred between the engineered and the biological systems.

As we move forward in developing biohybrid systems, we can leverage a second key trend in recent technological developments: technology is increasingly being designed to be adaptive in its capabilities. The breakthrough about to be achieved is to close the loop in a manner that utilizes the adaptive capabilities of electronic and mechatronic systems in order to promote adaptation in the nervous system.

1.2 NeuroDesign

The nervous system functions by generating patterns of neural activity. These patterns underlie sensation and perception as well as control of movement, cardiovascular, endocrine, immune, and other systems. Nonlinearities and dynamical states that span scales of physical form and time are key features of the patterns that emerge from the living nervous system. Biohybrid interfaces can be developed to (1) access these neural activity patterns, (2) influence the neural activity patterns, or (3) fundamentally alter the pattern formation mechanisms (i.e., promote plasticity) (Figure 1.1). This development can be accomplished through the process of “NeuroDesign.” One aspect of NeuroDesign is that the man-made abiotic systems to access or influence the neural patterns can be devised to embody the design principles of the nervous system. Here, the fundamental structure and/or operation of the technological system are based on an understanding of nervous system function. A second aspect of NeuroDesign is the process of engineering the nervous system itself. The concept here is a deliberate approach to mold and modify the structure and function of the nervous system to obtain a specific objective. In the short timescale, this can be thought of as “influence” or control of neural system function, in the medium timescale as “adaptation,” and in the long timescale as “plasticity or learning” of the nervous system. In closing the loop between the nonliving and the living, NeuroDesign also allows us to merge technology and science. This merger opens new opportunities for use of technological innovation for scientific investigation and a continuous modulation of biological activity to achieve desired function.

Figure 1.1 Biohybrid systems can access the patterns of neural activity, influence this pattern in real time, and induce plasticity by altering the pattern formation mechanisms. Brain image from http://www.getfreeimage.com/image/77/human-brain-and-neuron-impulses.

The primary challenge is to design biohybrid interfaces that can access and capture the biosignatures of the living system through limited spatiotemporal sampling and influence the inherently adaptive biological system through punctate intervention. For promoting plasticity, the challenge is to promote learning by influencing the core biochemical machinery in a desired manner.

1.3 The NeuroDesign Approach

Figures 1.2 and 1.3 illustrate the approach to NeuroDesign. The three features of this approach are (1) integration between the exogenous human designed system and the endogenous living system (2) biomimicry in the design of the exogenous system, and (3) the fact that an intervention that exerts its direct influence at one scale has an overall effect that spans multiple scales. The exogenous system performs both neurosensing and neuroactivation. By designing engineered systems that are biomimetic, we are able to produce systems with some of the robustness and versatility of biological systems and that potentially facilitate functional integration with the endogenous biological system. The nature and degree of biomimicry that could be used in the design of the exogenous system depend on the objective for which the biohybrid is developed. That is, when using a closed-loop system to discover ion channels at the cellular level, neuromimicry at the cellular level leads to utilization of computational models of neurons with details of ion channels. On the other hand, the development of systems for closed-loop rhythmic control of the neuromusculoskeletal system utilizes the concept of pattern generators in the nervous system to design the exogenous system.

Figure 1.2 “NeuroDesign” integrates man-made systems with biological systems to access information, influence the activation of the biological system in real time, and/or promote long-term plasticity in the biological system. Bidirectional communication at multiple points of interface offers opportunities for closed-loop control of coadaptive systems. Biomimetic approaches are often used in the design of the exogenous system.

Figure 1.3 Biohybrid interfaces between exogenous man-made systems and endogenous biological systems can occur at one or more junctions along multiple scales of form and complexity. The effects of the interface at any one scale are propagated along the chain of scales.

Biohybrid systems can effect outcomes at multiple scales, at the behavioral scale (function), electrophysiological scale (synaptic learning), morphological scale (form), or molecular scale (genes/proteins/sugars). An interface that acts at one scale influences the entire chain (Figure 1.3). Thus, changes brought about at the molecular microlevel affect the pattern of activation across scales and ultimately influence behavior on a macroscale. On the other end, intervention at the macroscale for, for example, electrical stimulation of peripheral nerves after incomplete spinal cord injury to provide repetitive movement therapy, can promote motor recovery perhaps by promoting neuroplasticity at the molecular level [1–4].

Biohybrid systems can thus facilitate investigation of the intact and diseased living systems to efficiently replace damaged biological systems and to effectively interact with the residual biological components with the promise of repair.

1.4 Neuromorphic Control of a Powered Orthosis for Crutch-Free Walking

The use of NeuroDesign in the deployment of biohybrid systems can be illustrated by the following example of a powered orthotic and prosthetic system that is driven by a neuromorphic controller that was designed using biomimetic NeuroDesign principles [5]. This biohybrid system (patent pending) is designed to allow “crutch-free” walking by a person with a tibial fracture of the lower limb. For this system, two objectives must be met: (1) the injured lower limb must be stabilized; and (2) the person must be able to walk under voluntary control. To achieve the former, the orthotic system illustrated in Figure 1.4 was designed. This device consists of a fixed-ankle orthosis that is used to stabilize or immobilize the injured lower limb. The fixed-ankle orthosis is encased by an actuated (powered) false-foot orthosis and the combined device forms an actuated articulated false-foot orthosis (AAFO). This AAFO is designed to permit the person to walk with a stabilized lower limb with minimal load bearing on the injured limb.

Figure 1.4 Prototype of a fixed universal ankle–foot orthosis (UAFO) attached to an AAFO. The prototype device is designed for use by combat troops. Quick release pins on the top and bottom can be used to easily separate the actuator from the AAFO.

In order to achieve the second objective and provide voluntary control of the false foot, it was necessary to access information about the intent of the person to walk and then appropriately control the cyclic movement of the AAFO during walking. The inspiration for the design of this control system scheme was drawn from the control of movement in biological systems. Networks of neurons in the spinal cord of vertebrates are capable of producing rhythmic neural output that in turn controls a well-orchestrated sequence of muscle activation for cyclic control of locomotion [6]. The activity of these spinal pattern generators is usually initiated and terminated by descending voluntary control signals from the brain. The pattern generators also receive feedback from sensors in actuated muscles and tendons during the entire gait cycle. The neural organization of this biological system was mimicked in the design of the control system used for the AAFO.

An electronic circuit was designed to implement a neural network pattern generator that could be used as the controller (Figure 1.5). The biomimetic architecture of the pattern generator circuit was based on knowledge of connectivity of neurons within the spinal cord of the lamprey, a primitive vertebrate [7, 8]. Computational models of individual neurons were implemented in a circuit made from analog very large scale integrated (aVLSI) components and discrete electronic components [9, 10]. This pattern generator is capable of autonomously generating cyclic voltage output that drives the AAFO. Biological pattern generators can be entrained by impinging cyclic rhythms. Their rhythm can also be reset if a perturbation of sufficient strength is applied at a particular phase of the rhythm. For example, the spinal pattern generator of the lamprey can be entrained by mechanosensory signals as well as reset by perturbations to stop and start anew [11]. Sensors mounted on the leg or AAFO provide cyclic input to the electronic pattern generator controlling the AAFO. In this manner, voluntary control of gait initiates and terminates cyclic actuation of the AAFO. Once initiated, the cadence of the AAFO matches the user's self-selected walking speed. Sensors mounted on the AAFO also provide two types of feedback signals. One set of signals feeds back position information to the actuator of the articulated ankle for local control, while another set of signals feeds information on external perturbation to the pattern generator and resets the cyclic control of the AAFO.

Figure 1.5 Biohybrid neuromorphic orthotic control system. The rhythmic movement sensor captures intent to move and provides periodic descending signals to entrain the unit pattern generator controller, which provides the cyclic voltage output needed to actuate the AFO. Sensors on the AFO provide local feedback (FBs) to the actuator for control of position and ankle stiffness; sensors also provide input to the pattern generator where it may reset the rhythm in the presence of perturbation (FBR).

The importance of having an actively controlled AAFO instead of just a passively controlled ankle–foot orthosis (AFO) becomes apparent during walking (Figure 1.6). When operating in passive mode (without active control), the false foot dorsiflexes during stance phase (at approximately 40% of the gait cycle) and does not actively plantar flex at the ankle during push-off (at approximately 60% of the gait cycle). With active control that is automatically timed by the entrained pattern generator, this dorsiflexion is prevented and the ankle more closely follows the normal ankle movement pattern.

Figure 1.6 Ankle kinematics (in degrees of dorsiflexion) during a typical normalized gait cycle (heel strike to heel strike) with no orthosis (normal), passive orthosis (left), and active AAFO (right). Active control of the orthosis corrects the excessive dorsiflexion during stance phase (at 40%) and provides more plantarflexion at push-off (at 60%).

Thus, this example shows how a neuromorphic design of a control system for a powered orthosis can function as a biohybrid device at the macroscale. It offers “crutch-free walking” to a person with an injured lower limb.

1.5 Frontiers of Biohybrid Systems

The greatest promise of biohybrid systems lies in promoting plasticity in the nervous system, thereby contributing to recovery and repair of lost biological function whether it ensues because of trauma, disease, or aging. This will be achieved as the closed loop becomes adaptive with adaptation occurring in both the biological and the engineered components. The greatest challenge is to design engineered systems whose adaptation enables the system to customize itself to each individual and to account for changes in the biological system as the two systems coadapt ([12–18].

As discussed and presented by multiple examples in this book, patterns of activity of the biological system could be accessed using advanced adaptive technology that responds to a biological system that is nonstationary and dynamic, and functions across multiple time- and spatial scales and multiple modalities. The design of the control system will be guided by the structural and functional constraints observed in biological systems, and allow for real-time learning, stability, and error correction that accounts for the biological systems features and takes into account the paucity of inputs to influence the biological system. The frontier lies in being able to harness the adaptive technology to promote plasticity and synergistic learning with the biological system on a long timescale under coadaptive conditions. Optimizing the technology will necessitate an approach that looks beyond the technology in isolation and looks beyond the technology as it interacts with the biological system in its current state. Here, the design of effective technology must consider its adaptive interaction with a biological system that is continuously changing.

Endogenous compensatory learning of the biological system on short and long timescales and the physical constraints of interaction will provide challenges to this synergistic learning. It is likely that there exist windows of opportunity that may be critical periods for induction of sustained learning. Learning in the merged systems will have occurred when there are carryover effects beyond the time period when the technology is interacting with the biological systems. Future biohybrid systems may have the ability to self-wean when necessary. The biohybrid systems will thus allow us to discover the principles governing activity-dependent learning in living systems, to develop novel approaches to sense the dynamic changes in the adaptive living system and the environment, and to deliver novel adaptive technology that encourages appropriate plasticity in biological systems.

1.6 Chapter Organization

The book chapters are divided into three sections. Together, the chapters illustrate the principle approaches of NeuroDesign and present practical applications of the use of biohybrid systems for scientific interrogation and medical intervention. The first three chapters present the principles that can be used for development of biohybrid systems. Chapter 2 presents the principles of computational neuroscience. Computation complements mathematical theory and is often used to understand and reengineer the neural code represented by the rich repertoire of neural activity patterns under natural as well as experimental conditions. This chapter introduces basic physiology of neurons and presents mathematical models for excitable cells. It also presents general formalisms in neuronal modeling and briefly captures models for plasticity. The ability to embody these equivalent mathematical models for neural cells and synapses in silicon using neuromorphic electronic design principles is presented in Chapter 3. Fundamental devices and circuits that can emulate neuronal behavior at the single cell level as well as more complex circuits are presented. The chapter also discusses the advantages of using a neuromorphic approach in the design of the hardware. Chapter 4 presents principles of signal processing. It specifically examines the use of point process theory for understanding the neural code and illustrates the bounds placed by this theory in the rational design of interfaces for biohybrid systems for neurosensing and neurostimulation.

The next three chapters discuss biohybrid systems that interface at the single cell level. Chapter 5 presents the role of dynamic clamp in biomimetic and biohybrid living-hardware systems. The concepts of the dynamic clamp experimental technique are discussed and illustrated. The technique utilizes artificial synapse interfaces between single cells and computational models of those cells to investigate the fundamental biochemistry of neuronal activation. Also presented are examples of use of such biohybrid systems for specific neuronal gain control by manipulating synapses. Approaches by which the actual interface between individual neurons and sensing transducers can be enhanced by surface modification of the hardware at nanoscales that mimic biology are presented in Chapter 6. This section wraps up with Chapter 7, which introduces real-time computing for the development of the artificial neurons utilized in dynamic clamp studies. It also presents an easy-to-learn and easy-to-use technique for performing biohybrid systems analysis and presents the use of a biohybrid system to control the heartbeat in a leech though dynamic clamp.

The last section of the book consists of four chapters on biohybrid systems that interface at a macroscale and present the potential for closed-loop control of complex systems using such interfaces. Chapter 8 on biomimetic adaptive control algorithms presents the use of biomimetic features including computational models of excitable neurons, network architectures derived from biological systems, and learning algorithms inspired by synaptic learning mechanisms for the design of adaptive control algorithms. The chapter also discusses factors that should be considered in the design of closed-loop control systems in the context of coadaptation of the interfaced systems. Chapter 9 builds on Chapter 3 by presenting applications that utilize neuromorphic hardware for audition and vision and a system to control the neuromuscular skeletal system after spinal cord injury. In Chapter 10, a new approach to control cardiac function by interfacing with the nervous system is presented. It discusses the precautionary measures that will be necessary in the design of a closed-loop system. Finally, a biohybrid system with an adaptive smart sensor to measure neural activity of pancreatic cells cultured on multielectrode arrays is presented in Chapter 11. The chapter also presents the initial building blocks for a closed-loop implantable system for measuring blood-borne glucose for the management of diabetes.

Acknowledgment

Neuromorphic controlled powered orthosis case study supported by STTR Contract No. W911NF-05-C-0122 from Army Research Office to Advensys, LLC (PI: R. Jung; Contributions t case study by S.V. Jung, V. Darbhe, B. Srimattirumalaparle, D. Meller and S. Phillips at Advensys LLC and J. J. Abbas, J. Jolly and T. Sugar at Arizona State University.).

References

1. Thuret, S., Moon, L.D.F., and Gage, F.H. (2006) Therapeutic interventions after spinal cord injury. Nat. Rev. Neurosci., 7, 628–643.

2. Lynskey, J., Belanger, A., and Jung, R. (2008) Activity dependent plasticity in spinal cord Injury. J. Rehabil. Res. Dev., 45 (2), 229–240.

3. Al-Majed, A.A., Neumann, C.M., Brushart, T.M., and Gordon, T. (2000) Brief electrical stimulation promotes the speed and accuracy of motor axonal regeneration. J. Neurosci., 20, 2602–2608.

4. Jung, R., Belanger, A., Kanchiku, T., Fairchild, M., and Abbas, J.J. (2009) Neuromuscular stimulation therapy after incomplete spinal cord injury promotes recovery of interlimb coordination during locomotion. J. Neural Eng., 6, 055010.

5. Jung, R., Jung, S.V., and Srimattirumalaparle, B. (2010) Neuromorphic controlled powered orthotic and prosthetic system. USPTO 201002806289 A1.

6. Gillner, S. (2006) Biological pattern generation: the cellular and computational logic of networks in motion. Neuron, 52, 751–766.

7. Grillner, S., Wallen, P., Brodin, L., and Lansner, A. (1991) Neuronal network generating locomotor behavior in lamprey: circuitry, transmitters, membrane properties, and simulation. Annu. Rev. Neurosci., 14, 169–199.

8. Buchanan, J.T. (1992) Neural network simulations of coupled locomotor oscillators in the lamprey spinal cord. Biol. Cybern., 66, 367–374.

9. Jung, R., Kiemel, T., and Cohen, A.H. (1996) Dynamic behavior of a neural network model of locomotor control in the lamprey. J. Neurophysiol., 75, 1074–1086.

10. Jung, R., Brauer, E.J., and Abbas, J.J. (2001) Real-time interaction between a neuromorphic electronic circuit and the spinal cord. IEEE Trans. Neural Syst. Rehabil. Eng., 9 (3), 319–326.

11. McClellan, A.D. and Jang, W. (1993) Mechanosensory inputs to the central pattern generators for locomotion in the lamprey spinal cord: resetting, entrainment, and computer modeling. J. Neurophysiol., 70 (6), 2442–2454.

12. Jung, R. (2007) Adaptive learning technology. Final Workshop Report on “Future Challenges for the Science and Engineering of Learning, National Science Foundation, USA, July 23–25, 2007” (http://www.nsf.gov/sbe/SLCWorkshopReportjan08.pdf).

13. Viadaurre, C., Sannelli, C., Muller, K.-R., and Blankertz, B. (2011) Machine-learning-based coadaptive calibration for brain–computer interfaces. Neural Comput., 23 (3), 791–816.

14. Gage, G.J., Ludwig, K.A., Otto, K.J., Ionides, E.L., and Kipke, D.R. (2005) Naive coadaptive cortical control. J. Neural Eng., 2, 52–63.

15. Wang, W., Collinger, J., Perez, M., Tyler-Kabara, E., Cohen, L., Birbaumer, N., Brose, S., Schwartz, A., Boninger, M., and Weber, D. (2010) Neural interface technology for rehabilitation: exploiting and promoting neuroplasticity. Phys. Med. Rehabil. Clin. North Am., 21 (1), 157–178.

16. Riess, J. and Abbas, J.J. (2001) Adaptive control of cyclic movements as muscles fatigue using functional neuromuscular stimulation. IEEE Trans. Neural Syst. Rehabil. Eng., 9, 326–330.

17. Dobkin, B.H. (2003) Do electrically stimulated sensory inputs and movements lead to long-term plasticity and rehabilitation gains? Curr. Opin. Neurol., 16, 685–691.

18. Fairchild, M., Kim, S.J., Iarkov, I., Abbas, J.J., and Jung, R. (2010) Repetitive hindlimb movement using intermittent adaptive neuromuscular electrical stimulation in an incomplete spinal injury rodent model. Exp. Neurol., 223, 623–633.

Chapter 2

Principles of Computational Neuroscience

Sharmila Venugopal, Sharon Crook, Malathi Srivatsan, and Ranu Jung

2.1 Introduction

The computational capacity of the nervous system is incredible. Computational neuroscience tries to capture the theoretical basis of this complex capacity at multiple scales using computational approaches. Computation complements mathematical theory and is often used to understand and reengineer the neural code represented by the rich repertoire of neural activity patterns under both natural and experimental conditions. A historical step in this direction dates back to the seminal works of Alan L. Hodgkin and Andrew F. Huxley who developed a neurophysiology-based mathematical model for the squid giant axon action potential [1]. Their formalism is extensively used even today, either in its original or in its reduced form. Another noteworthy contribution came from Wilfrid Rall, who emphasized the importance of the spatial domain and dendritic processing of synaptic input using mathematical approaches based on electrical cable theory [2–4]. Rall pioneered the use of digital computers in neuroscience employing a discretized version of cable theory [5], and this compartmental modeling has formed the basis for some of the most widely used software tools in computational neuroscience (e.g., GENESIS [6] and NEURON [7]). The term “computational neuroscience” was, however, coined only in the late 1980s by Eric L. Schwartz who organized a conference whose proceedings were later published in the form of a book, “Computational Neuroscience” [8]. Despite much controversy for a clear definition of the field, computational models of neuronal systems have nonetheless been successful in providing test beds for hypotheses and generating valuable neurobiological predictions. The realm of computational neuroscience has now extended beyond its original scope of complementing empirical findings and providing insights into the underlying neurophysiology. As discussed in later chapters in the book, computational neuroscience is now significantly expanding its role and influencing advancements in related fields of neuroengineering and neurotechnology. Several excellent textbooks and seminal articles have rigorously dealt with neurobiological principles and theories (e.g., [9–16]) as well as analytical tools and techniques used in computational neuroscience (e.g., [1, 2, 6–8, 17–22]). This chapter outlines some of the predominant physiological and theoretical principles guiding neuronal modeling further indicating some of the key applications of computational models for neuromorphic and biohybrid system design and development.

2.2 Some Physiology of Neurons

Neurons and their supporting glial cells form the principal cell types of the nervous system. The Neuron Doctrine, as developed after the works of neuroanatomists Camillo Golgi and Ramon y Cajal, describes neurons as the structural and functional units of the nervous system [23]. Neurons form the basic information processing units of the nervous system and typically have a basic architecture consisting of a cell body (soma), many input processes (dendrites), and an output process (axon) (see Figure 2.1). The soma contains the nucleus and much of the cell's machinery. Each neuron usually has a single axon that arises from the axon hillock of the cell body and extends away from the soma. The plasma membrane of the axon is called the axolemma and is specialized to conduct electrical impulses originating at the axon hillock away from the neuronal cell body to other neurons. Terminal arborizations of axons end in boutons that form close appositions with the dendrites, soma, or axons of other neurons. These junctions are called synapses. Axodendritic synapses between the axonal terminals of the presynaptic neuron and specialized structures called dendritic spines on dendrites of the postsynaptic neuron are most common. Communication between the presynaptic neuron and the postsynaptic neuron typically occurs through two types of synapses, electrical or chemical. An electrical synapse consists of a gap junction between the pre- and postsynaptic neurons that allows rapid direct flow of ionic currents between the neurons bidirectionally. In chemical synapses, the boutons at the axon terminal of the presynaptic neuron store synaptic vesicles containing neurotransmitter chemicals. The space between the presynaptic axon terminal and the apposing postsynaptic dendritic spine is called the synaptic cleft. Transmission of action potentials to the axon terminals of the presynaptic neuron leads to release of the neurotransmitter(s) into the synaptic cleft that consequently binds to ionotropic or metabotropic receptors (G-protein-coupled receptors) on the surface of the postsynaptic neuronal membrane. Transmitter binding to ionotropic receptors leads to opening of ion channels that result in the generation of an electrical signal in the postsynaptic neuron. Transmitter binding to G-protein-coupled receptors activates second-messenger pathways often resulting in long-term changes. Neurotransmitters can have an excitatory effect and stimulate the postsynaptic neuron to generate an electrical impulse. They can also have an inhibitory effect wherein the neurotransmitter may bind to its receptor on the postsynaptic neuron and reduce its ability to generate electrical impulses. Thus, neuronal communication and information exchange occur between neurons across synapses by means of electrical and chemical signaling.

Figure 2.1 Basic architecture of a typical neuron communicating with another neuron via a chemical synapse.

2.2.1 Membrane Potential

All living cells including neurons maintain a negative electrical gradient between the inside (cytoplasm) and the outside of the cells separated by their cell membrane. The cell membrane encompassing the cytoplasm is a lipid (fat) barrier separating ions (charged particles) in the intracellular and extracellular fluids. Ions can be positively (cation) or negatively (anion) charged. Physiologically, the extracellular space largely consists of sodium (Na+) and chloride (Cl−) ions and the intracellular space is rich in potassium (K+) and organic anions (A−) (see Figure 2.2).

Figure 2.2 Physiological distribution of prominent ionic species across the membrane. The organic anions (A−), potassium (K+) ions, sodium (Na+) ions, and chloride (Cl−) ions are shown along with directions of the force of diffusion due to concentration gradients and electrical gradients due to differential distribution of ions across the membrane. Larger font size implies higher concentration.

The restriction of these ions to their respective locations occurs due to electrochemical and physiological constraints. Ideally, the cell membrane is a perfect electrical insulator and hence the ions stay where they are. In particular, the larger organic anions stay in the intracellular space, conferring the neuron with a negative electrical gradient. Moreover, the differential distribution of individual ionic types across the membrane leads to their respective chemical concentration gradients. But the cell membrane consists of ion-selective channels made of complex proteins embedded within the membrane (e.g., potassium channels) that allow ions to flow across the neuronal membrane. For example, if K+ ions were free to diffuse out of the cell via open K+ channels down its concentration gradient, the negative electrical gradient of the cell prevents the outward movement of these cations. Similarly, the inward flow of the negatively charged Cl− ions down their concentration gradient is counterbalanced by the negative membrane potential. But for Na+ ions, both the electrical and the concentration gradients drive the Na+ ions into the cell that could eventually neutralize the negative electrical gradient. Fortunately, at least two mechanisms prevent the inward Na+ ion flow. The foremost is the sodium–potassium transporter (also a membrane protein, often called Na+/K+ pump). The transporter utilizes cellular energy reserves (ATP) and exchanges three Na+ ions for two K+ ions across the cell membrane assisting in maintaining higher Na+ concentration outside the cell. Second, in addition to the ion selectivity, most channels are sensitive to membrane potential (voltage sensitive). Hence, the voltage-sensitive Na+ channels are typically closed at negative membrane potentials further preventing Na+ entry into the cell.

The balance between the electrical and the chemical gradients results in equilibrium potentials for each ionic species. Thus, an equilibrium potential exists for the K+ ions at which the electrochemical gradients balance each other and no net movement of K+ ions occurs. This equilibrium potential for potassium ions can be estimated using the Nernst equation given by

(2.1)

where R is the gas constant, T is the absolute temperature and [K+]OUT and [K+]IN are the extracellular and intracellular potassium ion concentrations, respectively. Similarly, equilibrium potentials can be obtained for each ionic species. The combined effect of more than one ionic species on the membrane potential can be expressed by the Goldman–Katz equation as follows:

(2.2)

where Vm is the membrane potential and P is the relative membrane permeability for each ion type. This equation includes the contribution of any ionic gradient to the membrane potential by simply weighting its effect in accord with its membrane permeability. Note that this estimation of membrane potential is based on the assumption of constant electric field (or potential gradient) across the membrane. The resultant membrane potential is referred as the resting membrane potential or Vrest (e.g., −70 mV).

2.2.2 Membrane Equivalent Circuit

The neuronal membrane bears close correlation with electrical circuits. The lipid bilayer acts like a capacitor forming a thin insulating barrier for the ions in the intracellular and extracellular spaces. Let Vm(t) describe the membrane potential difference between the intracellular and the extracellular domains at any given instant of time. The capacitance of the cell membrane (Cm) is a measure of the charge (Q) distributed across the membrane to give rise to Vm given by . Current flows when the voltage across the capacitor changes; this capacitive current is given by

In addition, the existence of various ionic channels confers the ability to conduct charges. As noted earlier, a resting membrane displays a negative potential and resistance to the flow of charges via the ion channels. The ion-permeable channels therefore contribute to the membrane resistance, Rm. Considering a point representation of the neuron (ignoring the complex morphology), we can represent a resting membrane by means of a simple resistance–capacitance equivalent circuit with a battery Vrest as shown in Figure 2.3.

Figure 2.3 Equivalent circuit of the electrical properties of the neuronal membrane.

The total current flowing across the membrane (Im) is the sum of the resistive (Ii) and capacitive (Ic) components. Applying Kirchoff's current law, the total current can be given by

where Cm is the specific membrane capacitance (F/cm2) and Rm is the specific membrane resistance (Ω cm2), and Gm, the inverse of Rm is the specific membrane conductance (S/cm2). The equivalent circuit described in Figure 2.3 includes many constituent ionic currents described in Section 2.2.1. The term Ii is therefore a sum total of various individual ionic current types as follows:

Here, the current Ileak is due to the conductance Gm and is voltage independent. The general expression for any given ionic current is given by

where Gion(Vm(t),t) describes voltage-dependent ionic conductance for a given type of ionic channel; Eion is the ionic reversal potential given by Nernst's equation for a particular ionic species. The reversal potential is the membrane potential at which a given ionic current reverses polarity. The total membrane current, Im in the current balance equation is set to zero due to conservation of charge. Hence, the membrane current balance equation can be written as follows:

(2.3)

2.2.3 Action Potential: Generation and Propagation

It is possible to experimentally gain access to the cell membrane by means of sharp microelectrodes (≤2 µm diameter) and introduce artificial currents to alter the membrane's potential (current clamp). Alternatively, the membrane potential can be held constant at different voltages to record the voltage-sensitive gating properties of various ionic membrane currents (voltage clamp). These two approaches have been integral to our understanding of the membrane physiology and to the development of physiologically realistic computational models of single neurons and synaptic currents.

Generation of an action potential (see Figure 2.4) involves rapid movement of ions across the membrane producing a transient change in the membrane potential leading to all or none electrical events (also called nerve impulse, electric impulse, or spike). Following from Section 2.2.1, if we inject a direct current of polarity such that it would result in de polarization of the resting membrane potential to more positive values, the depolarization-sensitive Na+-selective ion channels open; Na+ ions flow into the cell along their electrochemical gradient in turn making the membrane potential more positive. Increasing the magnitude of current injection will eventually allow the membrane potential to cross a certain threshold, resulting in an action potential (a rapid event typically lasting 1–2 ms). During the action potential, the membrane potential rapidly increases to about 40 mV (approximately the reversal potential for Na+ ions) and Na+ channels become refractory and enter an inactive state. The positive membrane potential in turn activates a set of voltage-sensitive K+ channels called delayed rectifiers that allow K+ to flow out of the cell along its concentration gradient. The outflow of K+ ions often undershoots the membrane potential below resting level and the membrane potential slowly returns to Vrest leading to an after-hyperpolarization (AHP) following an AP (see Figure 2.4); the process underlies closing of the delayed rectifier K+ channels as Na+ channels are simultaneously deinactivated and become ready to be activated again. The refractoriness of Na+ channels makes the generation of a second action potential impossible until the membrane is polarized to negative values close to its resting potential. Furthermore, the Na+/K+ pump enables extrusion of Na+ ions that entered the cell during the action potential. Thus, the generation of action potential depends on threshold and is refractory. While the mechanisms of a classical Na+ impulse can be thus described, action potentials of different neurons have varied amplitudes, shapes, and durations largely owing to the enormous diversity of the underlying ionic channels and their mechanisms [24].

Figure 2.4 Generation of a typical action potential (AP).

The action potential generated at or near the soma is conducted along the axon with nearly no depreciation in its amplitude and duration. The axonal membrane resembles an electrical cable with passive electrotonic properties (known as cable properties; also see Section 2.3.3) with resistance and capacitance coupled along the length of the axon. The electrotonic currents at the site of initiation of the action potential spread along the neighboring membrane sites where, upon reaching threshold, the action potential is regenerated. The rate of spread of depolarization largely depends on the size of the axon (larger axons have higher conduction rates) (also see chapters in Refs [12, 19]).

2.3 General Formalisms in Neuronal Modeling

2.3.1 Conductance-Based Hodgkin–Huxley Model for Action Potential Generation

The seminal works of Alan Hodgkin and Andrew Huxley resulted in the first physiologically realistic mathematical model for the generation of neuronal spikes. Their model formulation was based on a series of ion replacement and voltage clamp experiments where they studied the voltage- and time dependencies of the ionic conductances underlying the action potential in a giant axon of the squid ([25–28]). Following the physiological description of action potential generation in Section 2.2.3, the Hodgkin–Huxley (HH) model consists of conductance-based formalism for the fast sodium and delayed rectifier potassium currents. In particular, they introduced fictitious gating variables that described the activation and inactivation of these currents in a voltage- and time-dependent manner. Following equation 2.3 in Section 2.2.2, the total membrane current in the HH model was given by

(2.4)

where Iinj(t