Ebook: Neuroelectrodynamics

The brain is an incredibly powerful computing machine shaped by an evolutionary process that has developed over millions of years. The essence of brain function consists in how information is processed, transferred and stored. Understanding how our brain computes can explain why our abilities of seeing and moving in space are substantially more powerful as compared to any artificially built system. Without a clear model of how computation is performed in the brain, we are not able to understand the relationship between perception learning, memory and molecular changes at DNA, gene or protein level. Many models in computational neuroscience claim the existence of parallel, distributed processing system in the brain. Similarly many advances in the structural, microscopic and genetic fields have shed light on the molecular machinery within the brain. Additionally, we know that the brain processes and communicates information through complex electrochemical processes. Any theoretical and computational model must be able to bring all of these aspects together and build a generally true model of brain computation. However, this has not happened.

From a neurophysiological point of view, current doctrine remains focused within a spike timing paradigm. This paradigm has limited capability for advancing the understanding of how the brain works. Additional hypotheses such spike stereotypy (a visual assumption), temporal coding (which does not explain information processing) and cable function (which cannot explain memory storage and simplifies a complex, dynamic axonal system into a transmission cable) are required, due to which the reach of the neural computational model is limited. Assumptions made in any theoretical construct that are unproven make the theory limited and thus inherently weak. Taken from this perspective any subsequent result and further hypotheses are invalid.

These reasons were the beginning of our dissatisfaction with the state of the current trend in neuroscience and neural computation and the main reason for writing this book. Our view is that information processing in the brain transcends the usual notion of action potential occurrence with millisecond precision. The experimental evidence and the models presented in this book extensively demonstrate, for the first time to our knowledge, that spike timing theory considerably restricts the richness of how the brain carries out computations. The ability to record voltage extracellulary directly implies the existence of significant changes in local electric fields. Indeed action potentials themselves are shaped by the electric field and the presence, distribution and dynamics of charges that locally change molecular dynamics and electrical excitability of voltage-gated ion channels in active neurons.

We therefore set out to put forward a new model from a ground-up set of universal principles on the basis established by the laws of physics. These laws such as those of classical mechanics, thermodynamics and quantum physics can be used as guiding principles based upon which a general theoretical construct of brain computational model can be built. Such model attempts to incorporate the neurobiology of the cells, the molecular machinery itself along with the electrical activity in neurons to explain experimental results and in addition predict organization of the system.

In this book we show that spike timing models are special cases of a broader theoretical construct – NeuroElectroDynamics (NED) neuro- from neurons, electro- from electric field and -dynamics meaning movement. The main idea of NED is that under the influence of electric fields, charges that interact perform computations and are capable to read, write and store information in their spatial distribution at molecular level within active neurons. In this domain, the universal physical laws from classical mechanics to thermodynamics and quantum mechanics are applied in order to give mathematical and theoretical support to charges as the operators of the brain computing machine. Experimental data is used where necessary to corroborate, support and extend these observations.

This novel and substantial approach of the NeuroElectroDynamic model extends the previous approach to a different level where the physical aspect of information processing is fundamental and can be directly related to macromolecular changes, thermodynamic properties and intrinsic quantum phenomena. This new model represents the crucial paradigm shift which opens the door to a new subfield of neuroscience that bridges neural computational models and research in molecular computation.

Any model requires certain assumptions and the proposed theory is no different. However, the assumptions that are made in this theory are those based upon already validated laws and physical principles that should and indeed must be transferable between the fields of physics and neuroscience. No simplifications to these laws are made in the NED construct. The very novelty and strength of the model is in such an application of basic physical principles which are used to explain how the brain computes. This makes the theoretical framework used to generate NED “general” and by default very amenable to expansion, with clear possibility for experimental verification.

The book is organized into five large chapters. The first chapter outlines and provides an overview of the deficiencies of the current approaches to neural computation. The current development of neuronal computation based upon the assumption of spike stereotypy, known as spike timing theory is fraught with several gaps and controversies. Many of these issues are then discussed in detail in individual chapters along with a parallel development of the new theoretical architecture alluded to in this first chapter. The chapter is also a gentle introduction to laws and principles required to build a new model that addresses the issues raised in the chapter. This description highlights a systemic approach where universal laws should shape the nature of computations.

Since the link between electrophysiological properties of neurons and theory of neural computation, is critical, Chapter 2 describes such experiments and their results. The analyses performed in the in-vivo behaving animals are presented with a very fresh approach to reflect the essential characteristics of neurons and neuronal ensembles and disqualify spike timing as a unique modality which can fully describe brain computations. Spike timing analyzed in local or larger areas does not seem to be rich enough and does not tell us the whole story. Fundamental processing is hidden, observables such the timing of spikes (firing rate, ISI), oscillations partially include information regarding these hidden processes. Importantly, new methods and techniques, designed to characterize neuronal activity well beyond the spike timing approach are revealed in this chapter. This new description provides the required link between computation and semantics analyzed during behavioral experiments.

Chapter 3 highlights some important concepts which relate the theory of information with computation and electrophysiological properties of neurons. The presentation differentiates between popular view of ciphering (i.e. performing encryption and decryption) and fundamental mathematical principles of coding and decoding information. Previous theories of neural coding are discussed in this context and new concepts are presented. This novel view relates information processing and computation with cognitive models. Information flow within the nervous system is driven by electrical phenomena with full support from the biological and neurochemical infrastructure of the neurons.

Chapter 4 is the heart of the presentation of the new model of computation and forms the raison d'être for the book. The chapter builds a systemic approach of brain computation in terms of interactions, dynamics of charges (NeuroElectroDynamics) from the nanoscale level to macroscale phenomena using a theoretical design, the computational cube, which can be regarded as a compartmental model. This novel model proves that complex, grained charge level of information processing is the subtle phenomenology behind the commonly observed spiking activity and temporal pattern occurrence. Such processes have implicit quantum mechanisms that can be demonstrated as intrinsic results of experimental data analysis. A unified mathematical view of computation expressed in terms of charge dynamics as successive coding and decoding phases as developed in the NED theory is able to characterize the activity within single cells or neuronal ensembles.

It is clear that computation depends on intrinsic physical laws (the principle of least action, second law of thermodynamics) and anatomical substrate of neurons and is reflected in their electrophysiological properties. The concepts of computation based on the dynamics and interaction of charges, is no different and is the grounding upon which NED model has been developed. Additionally, the new approach is able to explain “coding and decoding” mechanisms, information retrieval, perception, memory formation, and other cognitive processes. We show as an example that temporal organization/temporal patterns can be obtained as a result of ‘learning’ starting with a charge movement model. The description of these state-of-the-art techniques of information processing (coding and decoding) are gradually developed in this book, using a simple mathematical formalism. This approach will allow readers from different backgrounds to implement and experiment new models and perform computer simulations of brain computation. There is no real equivalent for such development in the models that consider temporal aspects of spike occurrence (firing rate, ISI). The theoretical approach of NED builds a mathematically general model of brain computation that incorporates the current spike timing theory as a particular case. This ability of generalization is sorely missing from the current models of brain function. As the NED theory is accepted and developed in the future, it is likely that it can itself be incorporated within another even more generalized approach.

Chapter 5 establishes and expounds the new NED model and several examples of its application are shown in order to reveal and characterize brain computations. These applications are only partially developed here in this book and will form the basis of our next work. Several processes including memory and cognition can be explained if spike timing models are replaced with the new systemic approach of dynamics and interaction of charges. The emergent topology of charges as a result of their intrinsic dynamics during action potentials and synaptic spikes are at the origin of distance-based semantics, a phenomenon previously experimentally evidenced in several brain recordings.

NeuroElectroDynamics is a completely different model where the intrinsic computational processes are described by the dynamics and interaction of charges. Such a description very naturally includes and incorporates the internal neuronal molecular processes, the effect of neurotransmitters and genetic regulatory mechanisms. At fundamental molecular level, information carried by electric field and charges can be selectively transferred and can perform intricate operations on regulatory mechanisms at the gene level which are required to build appropriate proteins. Importantly, NeuroElectroDynamics shows that computational properties of neurons are built and change at the quantum level. The subtle changes that occur at a molecular level determined by the interactions of charges and electric fields can be related with qualitative information processing. Most important to the concept of human behavior, it is possible to relate these interactions to the emergent property of semantics of behavior, which has been elusive to date except in the realms of psychology and psychophysics.

As presented this multimodal aspect of information transfer and processing in the brain incorporates and indeed requires interactions with intracellular neuronal and non-neuronal complex molecular regulatory mechanisms from gene selection/expression, DNA computations to membrane properties that spatially shape the dynamics and interaction of charges. Spike timing including spike patterns can be regarded as an epiphenomenon or a result of these hidden processes. The often observed heterogeneity in activation as response to a given stimulus reflects strongly processes that occur at molecular level within selected neurons.

Therefore, NeuroElectroDynamics provides the structure upon which one can bridge molecular mechanisms and the spike timing framework. This systemic view unifies neural computation and may provide fresh guidance to research in artificial intelligence using newly developed models of computation. Instead of revolving around spike timing code the new model highlights several different levels where information is processed, emphasizes and indeed necessitates within the theory the role of fine sensitive structure of cells, the role of ion channels, proteins, genes and electric charges in information processing.

Additionally, this chapter alludes to the role of afferent brain rhythms including sleep in learning and memory storage. Also, specific neuropathological situations where the brain is diseased such as in schizophrenia or Parkinson are briefly developed in the context of NED. The analysis of brain processes in terms of dynamic changes and resulting electric perturbations is fundamentally necessary to allow us to consider a true brain-computer interface or to perform what we feel is true neuromodulation. The future of these concepts and how we should achieve them are also discussed here. We explain why currently built artificial systems based on ultra-simplified neurons using spike timing paradigms are inconsistent with brain computations and their future is limited by this reductionist approach. Built on principles supported by established pillars of physical laws in describing brain computation NeuroElectroDynamics is a step forward, required to develop the next generation of artificial intelligent systems and understand the mind in computational terms. We anticipate that the development in nanotechnology will allow this progress to be made soon.

The NED model highlights the importance of charge dynamics, electric interactions and electric field in information communication at a nanoscale level of protein synthesis. Traditional molecular biological approaches are unable to link the electrophysiological components of neural function to protein and genes and indeed most importantly with information models. Although NED does not provide a complete solution, we feel that it at least gives a path along which further research could be carried out in a way that emphasizes the full relationship between genes, proteins and complex regulatory mechanisms of information flow.

We started this journey more than five years ago by demonstrating that spikes are not stereotyped events by computing spike directivity and providing electrical images of recorded spikes. Since the fundamentals of spike timing theory of computation in the brain are based on the assumptions of ‘spike stereotypy’, representation of axons as cables and others, the resulting temporal coding theory and its conclusions are restrictive, limited and possibly inaccurate. Indeed, they are certainly incomplete in terms of explaining the complexity of function and behavior that we see in the real world.

After we began our effort of writing this book in February 2008, terms that we have defined such as – spike timing code as “a myth” – have already been borrowed from our proposal to publishers, by other experts in the field. We are delighted to see that some distinguished members of the academic community are beginning to incorporate these ideas into their own work. We are fully aware of how slow progress is in science and how hard it is to induce a paradigm change (see Planck, 1949; Kuhn, 1962).

We hope that this book will appeal greatly to many students. The models hinted at in chapter 5 can serve as research pathways for both experimental and theoretical development. Those that study physics and mathematics will know that ideas put forward by giants in their fields such as those of Schrödinger and Sir Roger Penrose can indeed be applied in neuroscience. New tools and mathematical concepts will need to be developed and applied. The students that are interested in systems neuroscience can revel in the fact that complex systems dynamics can be made to work and deeply explain experimental data. To the neurophysiologist there are now clear testable conjectures that are predicted and will result in a new line of experiments well beyond the current spike timing measurements. To the computer science and AI group, a true and detailed framework for brain function will allow much better algorithmic development of computations that can be translated into applications. From philosophical, psychological or psychoanalytical point of view, mind and brain are inseparable and this model of computation will appear as a naturally significant change that resonates strongly with previous views and new concepts of the science of the mind. Finally, to biologists that are struggling to find connections between the myriads of proteins, genes, their properties and behavior, NeuroElectroDynamics can be way of making sense of their findings in a comprehensive way.

The list of scientists that pioneered this work of in-depth investigation of intricate phenomena in brain is long. We can mention here all the Nobel laureates cited for their contribution in this field along other names such as Wilder Penfield, Mircea Steriade, Alwyn Scott, Roger Penrose, Stuart Hameroff and others that did not receive this honorary recognition but made significant observations in analyzing how the brain works. By contrast there is an even larger list of scientists that have simplified their observations to the temporal domain by building spike timing models, emphasizing their role and creating the current fashion in computational neuroscience.

We hope that for a new generation of students, researchers and scientists an important legacy stands in novel concepts initiated and developed in this book. We anticipate that their further development would establish a new field of research to emerge that will determine a paradigm shift required to significantly change current insights in computational neuroscience and connected fields which finally will provide a better understanding of the human brain and mind.

Acknowledgements

We have succeeded in publishing this book in this environment where every model that does not display the fashion of spike timing is highly marginalized in the mainstream publications. We have to thank those people that had the vision and made this publication possible especially to reviewers and publisher Mark Eligh from IOS Press.

Our foremost acknowledgement is to Shrikrishna Jog for carefully correcting the manuscript for valuable insights and useful comments. Also, the authors would like to express many thanks to Dr. Radu Balan, Dr. Neil Lambert and Liviu Aur for several suggestions to improve this book. Finally, we would like to thank our parents and families especially Monica, Teodora and our daughters, Alexandra, Rachna and Mihika for their patience.

The Authors