Content-Length: 597743 | pFad | https://gatech.academia.edu/StevePotter

Steve M Potter | Georgia Institute of Technology - Academia.edu

Skip to main content

Steve M Potter

Georgia Institute of Technology, Biomedical Engineering, Faculty Member

Followers

268

Following

14

Co-authors

13

Public Views

NeuroEngineering and Teaching consultant; semi-retired brain scientist & professor. Research: Closed-loop neural interfaces; multi-electrode arrays; learning and memory mechanisms; in vitro neuronal networks; epilepsy; prosthetics; multi-photon microscopy; live cell imaging.
Real-world courses: Introductory Neuroscience BMED 4752; Neuroengineering Fundamentals BMED 4400; Hybrid Neural Microsystems BMED 8813.
Supervisors: Jerome Pine, Scott E. Fraser, and Dana W. Aswad
Phone: +1 404 385 2989
Address: 313 Ferst Dr.
m/c 0535
Atlanta, GA, USA
30332-0535

less

William P. Hall

University of Melbourne

University of Florida

Matthew Gladden

Georgetown University

University of Leeds

Ludovic A Krundel

Hong Kong Polytechnic University

The University of Western Australia

SWPS University of Social Sciences and Humanities

Martyna Michalska

The University of Lodz

Ryszard W . Kluszczyński

The University of Lodz

InterestsView All (9)

Uploads

Books by Steve M Potter

Curriculum Vitae

Closing the Loop Around Neural Systems

Closed-loop neurophysiology has been accelerated by recent software and hardware developments and... more Closed-loop neurophysiology has been accelerated by recent software and hardware developments and by the emergence of novel tools to control neuronal activity with spatial
and temporal precision, in which stimuli are delivered in real time based on recordings or behavior. Real-time stimulation feedback enables a wide range of innovative studies of information processing and plasticity in neuronal networks. This Research Topic e-Book comprises 16 Original Research Articles, seven Methods Articles, and seven Reviews, Mini-Reviews, and Perspectives, all peer-reviewed and published in Frontiers in Neural Circuits. The contributions deal with closed loop neurophysiology experiments at a variety of levels of neural circuit complexity. Some include modeling and theoretical analyses. New enabling technologies and techniques are described. Novel work is presented from experiments in vitro, in vivo, and in humans, along with their clinical and technological implications for improving the human condition.

What can Artificial Intelligence get from Neuroscience?

The human brain is the best example of intelligence known, with unsurpassed ability for complex, ... more The human brain is the best example of intelligence known, with unsurpassed ability for complex, real-time interaction with a dynamic world. AI researchers trying to imitate its remarkable functionality will benefit by learning more about neuroscience, and the differences between Natural and Artificial Intelligence. Steps that will allow AI researchers to pursue a more brain-inspired approach to AI are presented. A new approach that bridges AI and neuroscience is described, Embodied Cultured Networks. Hybrids of living neural tissue and robots, called hybrots, allow detailed investigation of neural network mechanisms that may inform future AI. The field of neuroscience will also benefit tremendously from advances in AI, to deal with their massive knowledge bases and help understand Natural Intelligence.
Keywords: Neurobiology, circular causality, embodied cultured networks, animats, multi-electrode arrays, neuromorphic, closed-loop processing, Ramon y Cajal, hybrot.

Closing the Loop: Stimulation Feedback Systems for Embodied MEA Cultures.

For certain key functions, simple animals use large identified neurons, such as the locust's Gian... more For certain key functions, simple animals use large identified neurons, such as the locust's Giant Motion Detector neuron (LGMD), which integrates visual information and triggers jumping (Gabbiani et al., 1999). By contrast, in the vertebrate nervous system, individual neurons are probably not important; each function is subserved by many nerve cells working in concert. However, we have little understanding of how single-unit activity combines to form the network-level processing that takes in sensory input, stores memories, and controls behavior. Cultured neuronal networks have provided us with much of our present understanding of ion channels, receptor molecules, and synaptic plasticity that may form the basis of learning and memory (Bi and Poo, 1998; Latham et al., 2000; Misgeld et al., 1998; Muller et al., 1992; Ramakers et al., 1991). To study the nervous system in vitro offers many advantages over in vivo approaches. In vitro systems are much more accessible to microscopic imaging and pharmacological manipulation than are intact animals. Recent developments in multi-electrode array (MEA) technology, including those described below, will enable researchers to answer questions not just at the single-neuron level, but at the network level. Most MEA research has involved recording the activity that cultured networks produce spontaneously, via up to 64 extracellular electrodes. While some studies also included electrical stimulation via the substrate electrodes, it was applied to only one or two of them at a time (Connolly et al., 1990; Fromherz and Stett, 1995; Gross et al., 1993; Jimbo and Kawana, 1992; Jimbo et al., 1998; Maeda et al., 1995; Oka et al., 1999; Regehr et al., 1989; Shahaf and Marom, 2001; Stoppini et al., 1997). We propose that in order to substantially advance our understanding of network dynamics, we need high-bandwidth (many neuron) communication in both directions, out of and into the network. This chapter describes technologies that allow recording and stimulation on every electrode of an MEA, and a new closed-loop paradigm that brings in vitro research into the behavioral realm.

Two-photon microscopy for 4D imaging of living neurons.

T wo-photon laser-scanning fluorescence microscopy (Denk et al. 1990) has made it possible to ima... more T wo-photon laser-scanning fluorescence microscopy (Denk et al. 1990) has made it possible to image neurons over 600 pm deep within a living slice or organism, with submicrometer resolution and in three dimensions (3D) for many hours without photodamage (Potter et al. 1996a). With true 4D microscopy (i.e., 3D with time), it is now feasible to capture neural development and synaptic plasticity in the act of happening, eliminating many uncertainties associated with between-animal comparisons of fixed-tissue specimens. By using pulsed IR laser light to excite fluorescent labels (e.g., dyes, fluorescent proteins, or endogenous fluorophores) that are normally excited by visible light, excitation is restricted to the focal plane, greatly reducing photobleaching and phototoxicity (see Chapter 7). The IR illumination is scattered less than visible light, allowing imaging 2-3 times deeper than with standard confocal microscopy. In addition, because the fluorescent signal emanates only from the focus of the scanning IR laser beam within the specimen, no confocal aperture is necessary to remove the out-of-focus signal. This means that two-photon microscopy has an inherently higher signal-to-noise ratio (SNR) compared with confocal microscopy. Excellent references for two-photon microscopy, and for labeling and imaging living specimens, include Denk et al. (1995) and Terasaki and Dailey (1995). In this chapter, I describe two-photon imaging hardware, pointing out potential pitfalls in microscope construction and operation. I also describe technical considerations for successful two-photon microscopy in two experimental systems: (1) 4D imaging of living neurons transplanted to cultured hippocampal slices from neonatal rats and (2) 4D imaging of dendritic spines within acute hippocampal slices from adult rats. Figure 1 shows the components of the imaging setup.

Removing some 'A' from AI: Embodied Cultured Networks.

We embodied networks of cultured biological neurons in simulation and in robotics. This is a new ... more We embodied networks of cultured biological neurons in simulation and in robotics. This is a new research paradigm to study learning, memory, and information processing in real time: the Neurally-Controlled Animat. Neural activity was subject to detailed electrical and optical observation using multi-electrode arrays and microscopy in order to access the neural correlates of animat behavior. Neurobiology has given inspiration to AI since the advent of the perceptron and consequent artificial neural networks, developed using local properties of individual neurons. We wish to continue this trend by studying the network processing of ensembles of living neurons that lead to higher-level cognition and intelligent behavior.

Papers by Steve M Potter

The future of computing and neural interfacing: Wetware-Hardware Hybrids

Future Now: Reconfiguring Reality, 2017

The Institute for the Future asked me to predict where the next decade or so will take us with we... more

Targeted Stimulation Using Differences in Activation Probability across the Strength–Duration Space

by Steve M Potter, Michelle Kuykendal, and Martha Grover

Electrical stimulation is ubiquitous as a method for activating neuronal tissue, but there is sti... more Electrical stimulation is ubiquitous as a method for activating neuronal tissue, but there is still significant room for advancement in the ability of these electrical devices to implement smart stimulus waveform design to more selectively target populations of neurons. The capability of a device to encode more complicated and precise messages to a neuronal network greatly increases if the stimulus input space is broadened to include variable shaped waveforms and multiple stimulating electrodes. The relationship between a stimulating electrode and the activated population is unknown; a priori. For that reason, the population of excitable neurons must be characterized in real-time and for every combination of stimulating electrodes and neuronal populations. Our automated experimental system allows investigation into the stimulus-evoked neuronal response to a current pulse using dissociated neuronal cultures grown atop microelectrode arrays (MEAs). The studies presented here demonstrate that differential activation is achievable between two neurons using either multiple stimulating electrodes or variable waveform shapes. By changing the aspect ratio of a rectangular current pulse; the stimulus activated neurons in the strength–duration (SD) waveform space with differing probabilities. Additionally, in the case when two neuronal activation curves intersect each other in the SD space; one neuron can be selectively activated with short-pulse-width; high-current stimuli while the other can be selectively activated with long-pulse-width; low-current stimuli. Exploring the capabilities and limitations of electrical stimulation allows for improvements to the delivery of stimulus pulses to activate neuronal populations. Many state-of-the-art research and clinical stimulation solutions, including those using a single microelectrode, can benefit from waveform design methods to improve stimulus efficacy. These findings have even greater import into multi-electrode systems because spatially distributed electrodes further enhance accessibility to differential neuronal activation.

Closed-Loop Characterization of Neuronal Activation Using Electrical Stimulation and Optical Imaging

by Steve M Potter and Martha Grover

We have developed a closed-loop, high-throughput system that applies electrical stimulation and o... more We have developed a closed-loop, high-throughput system that applies electrical stimulation and optical recording to facilitate the rapid characterization of extracellular, stimulus-evoked neuronal activity. In our system, a microelectrode array delivers current pulses to a dissociated neuronal culture treated with a calcium-sensitive fluorescent dye; automated real-time image processing of high-speed digital video identifies the neuronal response; and an optimized search routine alters the applied stimulus to achieve a targeted response. Action potentials are detected by measuring the post-stimulus, calcium-sensitive fluorescence at the neuronal somata. The system controller performs directed searches within the strength–duration (SD) stimulus-parameter space to build probabilistic neuronal activation curves. This closed-loop system reduces the number of stimuli needed to estimate the activation curves when compared to the more commonly used open-loop approach. This reduction allows the closed-loop system to probe the stimulus regions of interest in the multi-parameter waveform space with increased resolution. A sigmoid model was fit to the stimulus-evoked activation data in both current (strength) and pulse width (duration) parameter slices through the waveform space. The two-dimensional analysis results in a set of probability isoclines corresponding to each neuron–electrode pair. An SD threshold model was then fit to the isocline data. We demonstrate that a closed-loop methodology applied to our imaging and micro-stimulation system enables the study of neuronal excitation across a large parameter space.

A Device for Long-Term Perfusion, Imaging, and Electrical Interfacing of Brain Tissue In vitro

by Nathan Killian, Steve M Potter, and Varadraj Vernekar

Distributed microelectrode array (MEA) recordings from consistent, viable, ≥500 µm thick tissue p... more Distributed microelectrode array (MEA) recordings from consistent, viable, ≥500 µm thick tissue preparations over time periods from days to weeks may aid in studying a wide range of problems in neurobiology that require in vivo-like organotypic morphology. Existing tools for electrically interfacing with organotypic slices do not address necrosis that inevitably occurs within thick slices with limited diffusion of nutrients and gas, and limited removal of waste. We developed an integrated device that enables long-term maintenance of thick, functionally active, brain tissue models using interstitial perfusion and distributed recordings from thick sections of explanted tissue on a perforated multi-electrode array. This novel device allows for automated culturing, in situ imaging, and extracellular multi-electrode interfacing with brain slices, 3-D cell cultures, and potentially other tissue culture models. The device is economical, easy to assemble, and integrable with standard electrophysiology tools. We found that convective perfusion through the culture thickness provided a functional benefit to the preparations as firing rates were generally higher in perfused cultures compared to their respective unperfused controls. This work is a step toward the development of integrated tools for days-long experiments with more consistent, healthier, thicker, and functionally more active tissue cultures with built-in distributed electrophysiological recording and stimulation functionality. The results may be useful for the study of normal processes, pathological conditions, and drug screening strategies currently hindered by the limitations of acute (a few hours long) brain slice preparations.

Asynchronous Distributed Multielectrode Microstimulation Reduces Seizures in the Dorsal Tetanus Toxin Model of Temporal Lobe Epilepsy

Brain Stimulation, 2015

Electrical brain stimulation has shown promise for reducing seizures in drug-resistant epilepsy, ... more Electrical brain stimulation has shown promise for reducing seizures in drug-resistant epilepsy, but the electrical stimulation parameter space remains largely unexplored. New stimulation parameters, electrode types, and stimulation targets may be more effective in controlling seizures compared to currently available options. We hypothesized that a novel electrical stimulation approach involving distributed multielectrode microstimulation at the epileptic focus would reduce seizure frequency in the tetanus toxin model of temporal lobe epilepsy. We explored a distributed multielectrode microstimulation (DMM) approach in which electrical stimulation was delivered through 15 33-µm-diameter electrodes implanted at the epileptic focus (dorsal hippocampus) in the rat tetanus toxin model of temporal lobe epilepsy. We show that hippocampal theta (6-12 Hz brain oscillations) is decreased in this animal model during awake behaving conditions compared to control animals (p &amp;amp;amp;amp;amp;amp;amp;amp;amp;amp;lt; 10(-4)). DMM with biphasic, theta-range (6-12 Hz/electrode) pulses delivered asynchronously on the 15 microelectrodes was effective in reducing seizures by 46% (p &amp;amp;amp;amp;amp;amp;amp;amp;amp;amp;lt; 0.05). When theta pulses or sinusoidal stimulation was delivered synchronously and continuously on the 15 microelectrodes, or through a single macroelectrode, no effects on seizure frequency were observed. High frequency stimulation (&amp;amp;amp;amp;amp;amp;amp;amp;amp;amp;gt;16.66 Hz/per electrode), in contrast, had a tendency to increase seizure frequency. These results indicate that DMM could be a new effective approach to therapeutic brain stimulation for reducing seizures in epilepsy.

Multimicroelectrode Microstimulation for the Treatment of Focal Epilepsy in the Tetanus-Toxin Rat Model

Objective : Our lab has previously shown that microstimulation distributed over multiple microele... more Objective : Our lab has previously shown that microstimulation distributed over multiple microelectrodes can eliminate epileptiform bursting in dissociated cultures of neocortical rat neurons (Wagenaar et al., 2005). This stimulation is carried out at lower frequencies and voltages than contemporary deep brain stimulation (DBS), making it a potentially more efficient and effective treatment for refractory epilepsy. The present study sought to validate these results in a live animal model. Methods : Adult male Sprague-Dawley rats were injected with 25 ng of tetanus toxin in the right dorsal hippocampus to induce chronic, spontaneous seizures and interictal spikes. A 16-channel microwire array (33 µm wire diameter) was chronically implanted, with electrodes positioned in the cell layers of CA1 and CA3. Biphasic, current-controlled pulses (±10 µA) were delivered asynchronously to 12 of 16 electrodes in a randomized order, using a custom-built stimulator. Distributed stimulation was car...

Real-world neuroscience assignments for learning what matters in the Real World

A real-time optogenetic feedback controller to clamp network firing

Proper neuronal and synaptic maturation is dependent on electrical activity patterns that occur d... more Proper neuronal and synaptic maturation is dependent on electrical activity patterns that occur during the development of neuronal circuits . Because many forms of neural activation are causally tied to somatic spiking, deducing the independent role of spiking and other forms of neural activation on network development has been difficult. Multichannel neural interfaces, such as microelectrode arrays, provide many independent measurements of network activity. This allows accurate real-time estimation of ensemble features of network activity, such as the population firing rate. Here, we combine a real-time, multichannel electrophysiology system with optogenetic techniques for network activation and inhibition to create a feedback controller capable of clamping population activity to user-defined set points. This system forces firing rates in cultured cortical to be rapidly and precisely follow set-points covering nearly three orders of magnitude (from 0.02 to 15 Hz/neuron). This techn...

Microelectrode array recordings reveal activity perturbations that trigger homeostatic synaptic plasticity

Homeostatic plasticity is a set of mechanisms for maintaining appropriate levels of spiking activ... more Homeostatic plasticity is a set of mechanisms for maintaining appropriate levels of spiking activity in developing neural circuits. When spiking in a cultured cortical network was blocked for 2 days using tetrodotoxin (TTX), there was a compensatory increase in excitatory synaptic strength (scaling). Upward scaling of synaptic strength also occured when fast glutamatergic transmission was blocked using 6-cyano-7-nitroquinoxaline-2,3-dione (CNQX). While changes in synaptic strength following pharmacological perturbations have been well-described throughout the nervous system, network spiking activity is rarely monitored during these perturbations. In this study, we continuously recorded network spiking activity in cortical cultures using microelectrode arrays (MEAs) during a 2-day application of TTX or CNQX. Drug treatments were followed by whole cell recordings to assess excitatory synaptic strength. We found that TTX effectively abolished all spiking activity; meanwhile, CNQX initi...

Book flyer for 50 Years of AI

A removable multi-well chambering system for neuronal networks cultured on micro-electrode arrays

Neuroscience For AI chapter pre-print

The human brain is the best example of intelligence known, with unsurpassed ability for complex, ... more The human brain is the best example of intelligence known, with unsurpassed ability for complex, real-time interaction with a dynamic world. AI researchers trying to imitate its remarkable functionality will benefit by learning more about neuroscience, and the differences between Natural and Artificial Intelligence. Steps that will allow AI researchers to pursue a more braininspired approach to AI are presented. A new approach that bridges AI and neuroscience is described, Embodied Cultured Networks. Hybrids of living neural tissue and robots, called hybrots, allow detailed investigation of neural network mechanisms that may inform future AI. The field of neuroscience will also benefit tremendously from advances in AI, to deal with their massive knowledge bases and help understand Natural Intelligence.

WagenaarNadasdy&Potter2006reprint

What Can AI Get from Neuroscience?

Lecture Notes in Computer Science, 2007

The human brain is the best example of intelligence known, with unsurpassed ability for complex, ... more The human brain is the best example of intelligence known, with unsurpassed ability for complex, real-time interaction with a dynamic world. AI researchers trying to imitate its remarkable functionality will benefit by learning more about neuroscience, and the differences between Natural and Artificial Intelligence. Steps that will allow AI researchers to pursue a more braininspired approach to AI are presented. A new approach that bridges AI and neuroscience is described, Embodied Cultured Networks. Hybrids of living neural tissue and robots, called hybrots, allow detailed investigation of neural network mechanisms that may inform future AI. The field of neuroscience will also benefit tremendously from advances in AI, to deal with their massive knowledge bases and help understand Natural Intelligence.

Curriculum Vitae

Closing the Loop Around Neural Systems

Closed-loop neurophysiology has been accelerated by recent software and hardware developments and... more Closed-loop neurophysiology has been accelerated by recent software and hardware developments and by the emergence of novel tools to control neuronal activity with spatial
and temporal precision, in which stimuli are delivered in real time based on recordings or behavior. Real-time stimulation feedback enables a wide range of innovative studies of information processing and plasticity in neuronal networks. This Research Topic e-Book comprises 16 Original Research Articles, seven Methods Articles, and seven Reviews, Mini-Reviews, and Perspectives, all peer-reviewed and published in Frontiers in Neural Circuits. The contributions deal with closed loop neurophysiology experiments at a variety of levels of neural circuit complexity. Some include modeling and theoretical analyses. New enabling technologies and techniques are described. Novel work is presented from experiments in vitro, in vivo, and in humans, along with their clinical and technological implications for improving the human condition.

What can Artificial Intelligence get from Neuroscience?

The human brain is the best example of intelligence known, with unsurpassed ability for complex, ... more The human brain is the best example of intelligence known, with unsurpassed ability for complex, real-time interaction with a dynamic world. AI researchers trying to imitate its remarkable functionality will benefit by learning more about neuroscience, and the differences between Natural and Artificial Intelligence. Steps that will allow AI researchers to pursue a more brain-inspired approach to AI are presented. A new approach that bridges AI and neuroscience is described, Embodied Cultured Networks. Hybrids of living neural tissue and robots, called hybrots, allow detailed investigation of neural network mechanisms that may inform future AI. The field of neuroscience will also benefit tremendously from advances in AI, to deal with their massive knowledge bases and help understand Natural Intelligence.
Keywords: Neurobiology, circular causality, embodied cultured networks, animats, multi-electrode arrays, neuromorphic, closed-loop processing, Ramon y Cajal, hybrot.

Closing the Loop: Stimulation Feedback Systems for Embodied MEA Cultures.

For certain key functions, simple animals use large identified neurons, such as the locust's Gian... more For certain key functions, simple animals use large identified neurons, such as the locust's Giant Motion Detector neuron (LGMD), which integrates visual information and triggers jumping (Gabbiani et al., 1999). By contrast, in the vertebrate nervous system, individual neurons are probably not important; each function is subserved by many nerve cells working in concert. However, we have little understanding of how single-unit activity combines to form the network-level processing that takes in sensory input, stores memories, and controls behavior. Cultured neuronal networks have provided us with much of our present understanding of ion channels, receptor molecules, and synaptic plasticity that may form the basis of learning and memory (Bi and Poo, 1998; Latham et al., 2000; Misgeld et al., 1998; Muller et al., 1992; Ramakers et al., 1991). To study the nervous system in vitro offers many advantages over in vivo approaches. In vitro systems are much more accessible to microscopic imaging and pharmacological manipulation than are intact animals. Recent developments in multi-electrode array (MEA) technology, including those described below, will enable researchers to answer questions not just at the single-neuron level, but at the network level. Most MEA research has involved recording the activity that cultured networks produce spontaneously, via up to 64 extracellular electrodes. While some studies also included electrical stimulation via the substrate electrodes, it was applied to only one or two of them at a time (Connolly et al., 1990; Fromherz and Stett, 1995; Gross et al., 1993; Jimbo and Kawana, 1992; Jimbo et al., 1998; Maeda et al., 1995; Oka et al., 1999; Regehr et al., 1989; Shahaf and Marom, 2001; Stoppini et al., 1997). We propose that in order to substantially advance our understanding of network dynamics, we need high-bandwidth (many neuron) communication in both directions, out of and into the network. This chapter describes technologies that allow recording and stimulation on every electrode of an MEA, and a new closed-loop paradigm that brings in vitro research into the behavioral realm.

Two-photon microscopy for 4D imaging of living neurons.

T wo-photon laser-scanning fluorescence microscopy (Denk et al. 1990) has made it possible to ima... more T wo-photon laser-scanning fluorescence microscopy (Denk et al. 1990) has made it possible to image neurons over 600 pm deep within a living slice or organism, with submicrometer resolution and in three dimensions (3D) for many hours without photodamage (Potter et al. 1996a). With true 4D microscopy (i.e., 3D with time), it is now feasible to capture neural development and synaptic plasticity in the act of happening, eliminating many uncertainties associated with between-animal comparisons of fixed-tissue specimens. By using pulsed IR laser light to excite fluorescent labels (e.g., dyes, fluorescent proteins, or endogenous fluorophores) that are normally excited by visible light, excitation is restricted to the focal plane, greatly reducing photobleaching and phototoxicity (see Chapter 7). The IR illumination is scattered less than visible light, allowing imaging 2-3 times deeper than with standard confocal microscopy. In addition, because the fluorescent signal emanates only from the focus of the scanning IR laser beam within the specimen, no confocal aperture is necessary to remove the out-of-focus signal. This means that two-photon microscopy has an inherently higher signal-to-noise ratio (SNR) compared with confocal microscopy. Excellent references for two-photon microscopy, and for labeling and imaging living specimens, include Denk et al. (1995) and Terasaki and Dailey (1995). In this chapter, I describe two-photon imaging hardware, pointing out potential pitfalls in microscope construction and operation. I also describe technical considerations for successful two-photon microscopy in two experimental systems: (1) 4D imaging of living neurons transplanted to cultured hippocampal slices from neonatal rats and (2) 4D imaging of dendritic spines within acute hippocampal slices from adult rats. Figure 1 shows the components of the imaging setup.

Removing some 'A' from AI: Embodied Cultured Networks.

We embodied networks of cultured biological neurons in simulation and in robotics. This is a new ... more We embodied networks of cultured biological neurons in simulation and in robotics. This is a new research paradigm to study learning, memory, and information processing in real time: the Neurally-Controlled Animat. Neural activity was subject to detailed electrical and optical observation using multi-electrode arrays and microscopy in order to access the neural correlates of animat behavior. Neurobiology has given inspiration to AI since the advent of the perceptron and consequent artificial neural networks, developed using local properties of individual neurons. We wish to continue this trend by studying the network processing of ensembles of living neurons that lead to higher-level cognition and intelligent behavior.

The future of computing and neural interfacing: Wetware-Hardware Hybrids

Future Now: Reconfiguring Reality, 2017

The Institute for the Future asked me to predict where the next decade or so will take us with we... more

Targeted Stimulation Using Differences in Activation Probability across the Strength–Duration Space

by Steve M Potter, Michelle Kuykendal, and Martha Grover

Electrical stimulation is ubiquitous as a method for activating neuronal tissue, but there is sti... more Electrical stimulation is ubiquitous as a method for activating neuronal tissue, but there is still significant room for advancement in the ability of these electrical devices to implement smart stimulus waveform design to more selectively target populations of neurons. The capability of a device to encode more complicated and precise messages to a neuronal network greatly increases if the stimulus input space is broadened to include variable shaped waveforms and multiple stimulating electrodes. The relationship between a stimulating electrode and the activated population is unknown; a priori. For that reason, the population of excitable neurons must be characterized in real-time and for every combination of stimulating electrodes and neuronal populations. Our automated experimental system allows investigation into the stimulus-evoked neuronal response to a current pulse using dissociated neuronal cultures grown atop microelectrode arrays (MEAs). The studies presented here demonstrate that differential activation is achievable between two neurons using either multiple stimulating electrodes or variable waveform shapes. By changing the aspect ratio of a rectangular current pulse; the stimulus activated neurons in the strength–duration (SD) waveform space with differing probabilities. Additionally, in the case when two neuronal activation curves intersect each other in the SD space; one neuron can be selectively activated with short-pulse-width; high-current stimuli while the other can be selectively activated with long-pulse-width; low-current stimuli. Exploring the capabilities and limitations of electrical stimulation allows for improvements to the delivery of stimulus pulses to activate neuronal populations. Many state-of-the-art research and clinical stimulation solutions, including those using a single microelectrode, can benefit from waveform design methods to improve stimulus efficacy. These findings have even greater import into multi-electrode systems because spatially distributed electrodes further enhance accessibility to differential neuronal activation.

Closed-Loop Characterization of Neuronal Activation Using Electrical Stimulation and Optical Imaging

by Steve M Potter and Martha Grover

We have developed a closed-loop, high-throughput system that applies electrical stimulation and o... more We have developed a closed-loop, high-throughput system that applies electrical stimulation and optical recording to facilitate the rapid characterization of extracellular, stimulus-evoked neuronal activity. In our system, a microelectrode array delivers current pulses to a dissociated neuronal culture treated with a calcium-sensitive fluorescent dye; automated real-time image processing of high-speed digital video identifies the neuronal response; and an optimized search routine alters the applied stimulus to achieve a targeted response. Action potentials are detected by measuring the post-stimulus, calcium-sensitive fluorescence at the neuronal somata. The system controller performs directed searches within the strength–duration (SD) stimulus-parameter space to build probabilistic neuronal activation curves. This closed-loop system reduces the number of stimuli needed to estimate the activation curves when compared to the more commonly used open-loop approach. This reduction allows the closed-loop system to probe the stimulus regions of interest in the multi-parameter waveform space with increased resolution. A sigmoid model was fit to the stimulus-evoked activation data in both current (strength) and pulse width (duration) parameter slices through the waveform space. The two-dimensional analysis results in a set of probability isoclines corresponding to each neuron–electrode pair. An SD threshold model was then fit to the isocline data. We demonstrate that a closed-loop methodology applied to our imaging and micro-stimulation system enables the study of neuronal excitation across a large parameter space.

A Device for Long-Term Perfusion, Imaging, and Electrical Interfacing of Brain Tissue In vitro

by Nathan Killian, Steve M Potter, and Varadraj Vernekar

Distributed microelectrode array (MEA) recordings from consistent, viable, ≥500 µm thick tissue p... more Distributed microelectrode array (MEA) recordings from consistent, viable, ≥500 µm thick tissue preparations over time periods from days to weeks may aid in studying a wide range of problems in neurobiology that require in vivo-like organotypic morphology. Existing tools for electrically interfacing with organotypic slices do not address necrosis that inevitably occurs within thick slices with limited diffusion of nutrients and gas, and limited removal of waste. We developed an integrated device that enables long-term maintenance of thick, functionally active, brain tissue models using interstitial perfusion and distributed recordings from thick sections of explanted tissue on a perforated multi-electrode array. This novel device allows for automated culturing, in situ imaging, and extracellular multi-electrode interfacing with brain slices, 3-D cell cultures, and potentially other tissue culture models. The device is economical, easy to assemble, and integrable with standard electrophysiology tools. We found that convective perfusion through the culture thickness provided a functional benefit to the preparations as firing rates were generally higher in perfused cultures compared to their respective unperfused controls. This work is a step toward the development of integrated tools for days-long experiments with more consistent, healthier, thicker, and functionally more active tissue cultures with built-in distributed electrophysiological recording and stimulation functionality. The results may be useful for the study of normal processes, pathological conditions, and drug screening strategies currently hindered by the limitations of acute (a few hours long) brain slice preparations.

Asynchronous Distributed Multielectrode Microstimulation Reduces Seizures in the Dorsal Tetanus Toxin Model of Temporal Lobe Epilepsy

Brain Stimulation, 2015

Electrical brain stimulation has shown promise for reducing seizures in drug-resistant epilepsy, ... more Electrical brain stimulation has shown promise for reducing seizures in drug-resistant epilepsy, but the electrical stimulation parameter space remains largely unexplored. New stimulation parameters, electrode types, and stimulation targets may be more effective in controlling seizures compared to currently available options. We hypothesized that a novel electrical stimulation approach involving distributed multielectrode microstimulation at the epileptic focus would reduce seizure frequency in the tetanus toxin model of temporal lobe epilepsy. We explored a distributed multielectrode microstimulation (DMM) approach in which electrical stimulation was delivered through 15 33-µm-diameter electrodes implanted at the epileptic focus (dorsal hippocampus) in the rat tetanus toxin model of temporal lobe epilepsy. We show that hippocampal theta (6-12 Hz brain oscillations) is decreased in this animal model during awake behaving conditions compared to control animals (p &amp;amp;amp;amp;amp;amp;amp;amp;amp;amp;lt; 10(-4)). DMM with biphasic, theta-range (6-12 Hz/electrode) pulses delivered asynchronously on the 15 microelectrodes was effective in reducing seizures by 46% (p &amp;amp;amp;amp;amp;amp;amp;amp;amp;amp;lt; 0.05). When theta pulses or sinusoidal stimulation was delivered synchronously and continuously on the 15 microelectrodes, or through a single macroelectrode, no effects on seizure frequency were observed. High frequency stimulation (&amp;amp;amp;amp;amp;amp;amp;amp;amp;amp;gt;16.66 Hz/per electrode), in contrast, had a tendency to increase seizure frequency. These results indicate that DMM could be a new effective approach to therapeutic brain stimulation for reducing seizures in epilepsy.

Multimicroelectrode Microstimulation for the Treatment of Focal Epilepsy in the Tetanus-Toxin Rat Model

Objective : Our lab has previously shown that microstimulation distributed over multiple microele... more Objective : Our lab has previously shown that microstimulation distributed over multiple microelectrodes can eliminate epileptiform bursting in dissociated cultures of neocortical rat neurons (Wagenaar et al., 2005). This stimulation is carried out at lower frequencies and voltages than contemporary deep brain stimulation (DBS), making it a potentially more efficient and effective treatment for refractory epilepsy. The present study sought to validate these results in a live animal model. Methods : Adult male Sprague-Dawley rats were injected with 25 ng of tetanus toxin in the right dorsal hippocampus to induce chronic, spontaneous seizures and interictal spikes. A 16-channel microwire array (33 µm wire diameter) was chronically implanted, with electrodes positioned in the cell layers of CA1 and CA3. Biphasic, current-controlled pulses (±10 µA) were delivered asynchronously to 12 of 16 electrodes in a randomized order, using a custom-built stimulator. Distributed stimulation was car...

Real-world neuroscience assignments for learning what matters in the Real World

A real-time optogenetic feedback controller to clamp network firing

Proper neuronal and synaptic maturation is dependent on electrical activity patterns that occur d... more Proper neuronal and synaptic maturation is dependent on electrical activity patterns that occur during the development of neuronal circuits . Because many forms of neural activation are causally tied to somatic spiking, deducing the independent role of spiking and other forms of neural activation on network development has been difficult. Multichannel neural interfaces, such as microelectrode arrays, provide many independent measurements of network activity. This allows accurate real-time estimation of ensemble features of network activity, such as the population firing rate. Here, we combine a real-time, multichannel electrophysiology system with optogenetic techniques for network activation and inhibition to create a feedback controller capable of clamping population activity to user-defined set points. This system forces firing rates in cultured cortical to be rapidly and precisely follow set-points covering nearly three orders of magnitude (from 0.02 to 15 Hz/neuron). This techn...

Microelectrode array recordings reveal activity perturbations that trigger homeostatic synaptic plasticity

Homeostatic plasticity is a set of mechanisms for maintaining appropriate levels of spiking activ... more Homeostatic plasticity is a set of mechanisms for maintaining appropriate levels of spiking activity in developing neural circuits. When spiking in a cultured cortical network was blocked for 2 days using tetrodotoxin (TTX), there was a compensatory increase in excitatory synaptic strength (scaling). Upward scaling of synaptic strength also occured when fast glutamatergic transmission was blocked using 6-cyano-7-nitroquinoxaline-2,3-dione (CNQX). While changes in synaptic strength following pharmacological perturbations have been well-described throughout the nervous system, network spiking activity is rarely monitored during these perturbations. In this study, we continuously recorded network spiking activity in cortical cultures using microelectrode arrays (MEAs) during a 2-day application of TTX or CNQX. Drug treatments were followed by whole cell recordings to assess excitatory synaptic strength. We found that TTX effectively abolished all spiking activity; meanwhile, CNQX initi...

Book flyer for 50 Years of AI

A removable multi-well chambering system for neuronal networks cultured on micro-electrode arrays

Neuroscience For AI chapter pre-print

The human brain is the best example of intelligence known, with unsurpassed ability for complex, ... more The human brain is the best example of intelligence known, with unsurpassed ability for complex, real-time interaction with a dynamic world. AI researchers trying to imitate its remarkable functionality will benefit by learning more about neuroscience, and the differences between Natural and Artificial Intelligence. Steps that will allow AI researchers to pursue a more braininspired approach to AI are presented. A new approach that bridges AI and neuroscience is described, Embodied Cultured Networks. Hybrids of living neural tissue and robots, called hybrots, allow detailed investigation of neural network mechanisms that may inform future AI. The field of neuroscience will also benefit tremendously from advances in AI, to deal with their massive knowledge bases and help understand Natural Intelligence.

WagenaarNadasdy&Potter2006reprint

What Can AI Get from Neuroscience?

Lecture Notes in Computer Science, 2007

The human brain is the best example of intelligence known, with unsurpassed ability for complex, ... more The human brain is the best example of intelligence known, with unsurpassed ability for complex, real-time interaction with a dynamic world. AI researchers trying to imitate its remarkable functionality will benefit by learning more about neuroscience, and the differences between Natural and Artificial Intelligence. Steps that will allow AI researchers to pursue a more braininspired approach to AI are presented. A new approach that bridges AI and neuroscience is described, Embodied Cultured Networks. Hybrids of living neural tissue and robots, called hybrots, allow detailed investigation of neural network mechanisms that may inform future AI. The field of neuroscience will also benefit tremendously from advances in AI, to deal with their massive knowledge bases and help understand Natural Intelligence.

The utility of intersitial, microfluidic perfusion in extended culturing of thick organotypic brain slices

Multi-electrode impedance tuning: The effect of electrode impedance on stimulation of dissociated cultures

Common median referencing for improved action potential detection with multielectrode arrays

Conference proceedings : ... Annual International Conference of the IEEE Engineering in Medicine and Biology Society. IEEE Engineering in Medicine and Biology Society. Annual Conference, 2009

Referencing is frequently used to remove common-mode signals from multielectrode data, in both fr... more Referencing is frequently used to remove common-mode signals from multielectrode data, in both freely moving animals and in vitro preparations. For action potential (AP) detection, referencing by subtracting the common average signal has been shown to increase AP signal-to-noise ratio (SNR). This method fails, however, when large transients occur on individual electrodes, as occurs during electrical stimulation or with large APs during spontaneous recordings. To deal with these cases, we propose using the common median as a reference. The common median has an improved SNR for AP detection (leading to more isolated single units and more detected APs per unit) and, unlike common average referencing, does not generate spurious APs when processing large single-electrode transients.

NeuroRighter: closed-loop multielectrode stimulation and recording for freely moving animals and cell cultures

Conference proceedings : ... Annual International Conference of the IEEE Engineering in Medicine and Biology Society. IEEE Engineering in Medicine and Biology Society. Annual Conference, 2009

Closed-loop systems, where neural signals are used to control electrical stimulation, show promis... more Closed-loop systems, where neural signals are used to control electrical stimulation, show promise as powerful experimental platforms and nuanced clinical therapies. To increase the availability, affordability, and usability of these devices, we have created a flexible open source system capable of simultaneous stimulation and recording from multiple electrodes. The system is versatile, functioning with both freely moving animals and in vitro preparations. Current- and voltage-controlled stimulation waveforms with 1 micros resolution can be delivered to any electrode of an array. Stimulation sequences can be preprogrammed or triggered by ongoing neural activity, such as action potentials (APs) or local field potentials (LFPs). Recovery from artifact is rapid, allowing the detection of APs within 1 ms of stimulus offset. Since the stimulation subsystem provides simultaneous current/voltage monitoring, electrode impedance spectra can be calculated in real time. A sample closed-loop ex...

Optogenetic feedback control decouples network spiking from other forms of neural activity

ABSTRACT Proper neuronal and synaptic maturation is dependent on electrical activity patterns tha... more ABSTRACT Proper neuronal and synaptic maturation is dependent on electrical activity patterns that occur during development. Because many forms of neural activation are causally tied to somatic spiking, deducing the independent role of spiking versus other activity types in network development has been difficult. Previous studies have demonstrated the power of optogenetic proteins for activating and/or inhibiting spiking activity in genetically specified neuronal populations using light. Here, we incorporate optogenetic interfacing techniques into a feedback control architecture in which multi-neuron spiking data is monitored and used to adjust an optical control signal to clamp population firing at a specified set-point. This &quot;optogenetic population clamp&quot; allows precise manipulation of firing levels in cultured cortical networks over many hours and can be used in conjunction with pharmacological techniques that disrupt other forms of neural activity, such as neurotransmission. This technique offers advantages in comparison with feedback control schemes using electrical stimulation as an input variable. Specifically, it provides a powerful input signal for more effective control of network recruitment, it is stable of over many hours, it can exert control over a large range of firing rates, and it provides a genetically specified input channel to complex, heterogeneous networks. Additionally, we show that using optogenetic feedback, network firing rates can be decoupled from perturbations to AMPA-receptor-mediated gluatametergic neurotransmission, which would normally have a large effect on spiking activity. This demonstrates the ability of this system to deconstruct neuronal networks into independently manipulable variables that would normally be causally intertwined. Therefore, this technique has important implications for research aimed at understanding the independent role of different forms of neural activity in long time-scale developmental processes.

Reservoir-computing-based, biologically-inspired artificial neural network for modeling of a single machine infinite bus power system

The 2012 International Joint Conference on Neural Networks (IJCNN), 2012

Abstract Inspired by living neuron networks (LNNs) in the brain, artificial neural networks (ANNs... more Abstract Inspired by living neuron networks (LNNs) in the brain, artificial neural networks (ANNs) have been broadly used in various applications as a computational intelligence tool. However, due to many fundamental differences between ANNs and LNNs, despite the mature training mechanisms for ANNs, it is often challenging to use LNNs as a computational intelligence tool. To bridge the gap between ANNs and LNNs, a novel type of artificial neural network, ie biologically-inspired artificial neural network (BIANN) is ...

NeuroEngineering: Neuroscience -- APPLIED

I define "neuroengineering" and provide some examples of its successes and where I think it is go... more

BETTER MINDS: Cognitive Enhancement in the 21st Century

BceM HaM XOLieTCR WMeTb 6onee COBepweHHbiCi, 6onee BOC npWWMLIWBbiCi pa3yM. nocKOnbKY 6wonorwYeCK... more BceM HaM XOLieTCR WMeTb 6onee COBepweHHbiCi, 6onee BOC npWWMLIWBbiCi pa3yM. nocKOnbKY 6wonorwYeCKWCi em HOCt-1-Ten b, M031, npWHWMaeT yYaCTwe BO BCeM, LITO Mbl ):lenaeM, eCTb BenWKOe MHO>KeCTBO npW-IWH, no KOTOpbiM Mbl MOint-1 6bl >Ke naTb ynyYweHwR Tom, c YeM pO>K):leHbl. MHorwe xoTenw 6bl TOL!Hee nOMHWTb BCRKWe Bel. ljt-1 -OT TpWBWan bHOIO "Ky):la R nonO>KWn(a) KniOYW?" ):10 Ba>KHOIO "a KOI):la y HaC IO):!OBI.ljWHa csa):lb6blr Mbl He npOYb 6b1Tb 6blcrpee w coo6pa3wTen bHee s wwpoKOM cneKTpe KOHTeKCTOB. BepomHo Mbl, niO):lw, pa3swnw CBOW He06bll . laCiHO BbiCOKWe KOIHWTWBHble CnOC06HOCTW, LIT0-6bl noMOLib ce6e pewaTb npo6neMbl, ycnewHo cnpasnmbcR C HeB310):laMW B ):!OWCTOpWL!eCKYIO 3noxy. 6blnO 6bl 3):!0pOBO, ecnw 6bl Haww KOIHWTWBHble HaBbiKW wnw sposeHb c nepeMe HaMw B >Kt-13Ht-1 niO):leCi, HO OHt-1 He W):lyT. >f<WBR B COBpeMeHHOM MWpe, Mbl pO>K):laeMCR C TeMW >Ke M031aMW, KaKWMt-1 o6na):lanw Haww ):!WKwe npe):!Kt-1 30 000 neT Ha3a):l. Tpy):! HOCTW, KOTOpble HaM npwxo):!WTCR npeo):lonesaTb ce IO):!HR, 3aMeTHO OTnWYaiOTCR OT npo6neM KpOMaHbOHI.. jeB. Mano KOMY Heo6xo):lwMo 3HaTb, KaK HaCiTw BO):ly, ):!06b1Tb nwlljy, cnaCTWCb OT XWI.ljHWKOB t-1 pa3BeCTW OIOHb. BMeCTO 3TOIO Haww MbiCnt-1 3aHRTbl TeM, KaKyiO Kapbepy W36paTb, KaK nona):!WTb C COCe):!RMt-1, KOTO pble BbllnR):!RT WHalle, LleM Mbl, l):le >Kt-'!Tb t-1 Ka KYIO KynwTb MawwHy. Mbl 6ecnoKOWMCR He o TOM, R): IOBWTO nw paCTeHWe, KOTOpOe Mbl C06wpaeMCR CbeCTb, a 0 TOM, ):!OCTaTOLI HO nw B HeM >Kene3a t-1 BWTaMWHOB. B OTCYTCTBWe >KWBOTHbiX, 3a KOTOpbiMt-1 Wnt-1 OT KOTOpbiX HY>KHO 6eraTb, t-1 npw W36biTKe Ka nopwCiHOCi nwlljw Hac 3a6oTWT, Ka K coxpaHWTb xopowy10 cpop MY B MWpe, He Tpe6yiOI.ljeM OT HaC 60nbWOIO HanpR>KeHWR cwn. HeKOTOpble t-13 "senwYaCiwwx npo6neM" Mwpa -3TO Ha ca MOM ):\ene He Te, 0 KOTOpbiX Mbl 06bll . IHO ):lyMaeM. 3acyxa t-1 10-nO):l -npo6neMa? nwl. ljt-1 t-1 BO): Ibl XBaTWT Ha BCeX, npOCTO OHt-1 He pacnpe):lenRIOTCR cnpase):lnv1Bo wnw 3cpcpeKTWBHO. 3HeprwR -STEVE M. POTTER BETTER MINDS: COGNITIVE ENHANCEMENT IN THE 21st CENTURY www.neuro.gatech.edu