Building Better Brains

Agenda

* Presentation times are subject to change.

Day 2: Friday, September 24, 2010

Session I: Basic Research with Strong Potential for Translation

Cognitive Neural Prosthetics

Richard Andersen, PhD
California Institute of Technology, Pasadena, CA

We are developing a neural prosthetic to assist paralyzed patients and patients with lost limbs by using their neural activity to control assistive devices such as computers, robotic limbs, and vehicles. Most efforts in this field have focused on recording neural activity close to the motor output. Our lab has taken a different approach which is to record from higher cognitive areas that form the initial intentions of the subjects. Using this approach in healthy animals we can decode the goals and trajectories of desired movements in space, the subjective evaluations and expectations that form the basis of decision making, and arbitrary visual-motor mapping. Current efforts include transitioning to clinical trials in humans.

A Neural Interface for Dexterity

Andrew B. Schwartz, PhD
University of Pittsburgh, Pittsburgh, PA

The emphasis on neural populations as the substrate for information processing is the most important recent advance in systems neuroscience. The change in emphasis from the single neuron to the neural ensemble has made it possible to extract high-fidelity information about movements that will occur in the near future. This ability is due to the distributed nature of information processing in the brain. Neurons encode many parameters simultaneously, but the fidelity of encoding at the level of individual neurons is weak. Because encoding is redundant— parameter representation in individual neurons is weak but consistent across the population-- extraction methods based on multiple neurons are capable of generating a faithful representation of intended movement. The realization that useful information is embedded in the population has spawned the current success of brain-controlled interfaces. Since multiple movement parameters are encoded simultaneously in the same population of neurons, we have been gradually increasing the degrees of freedom (DOF) that a subject can control through the interface. Our early work showed that 3-dimensions could be controlled in a virtual reality task. We then demonstrated control of an anthropomorphic physical device with 4 DOF in a self-feeding task. Currently, monkeys in our laboratory are using this interface to control a 7-DOF arm, wrist and hand to grasp objects in different locations and orientations. Our recent data show that we can extract 11-DOF to add hand shape and dexterity to our control set.

Applications of Recurrent Brain-Computer Interfaces

Eberhard Fetz, PhD¹, Andrew Jackson, PhD², Chet Moritz, PhD¹, Yukio Nishimura, PhD^1,3 Timothy Lucas, MD, PhD¹, and Steve Perlmutter, PhD¹

¹ University of Washington, Seattle, WA
² Newcastle University, Newcastle-upon-Tyne, United Kingdom
³ PRESTO, Japan Science and Technology Agency, Tokyo, Japan

We are investigating the consequences of bidirectional connections produced by an autonomous recurrent brain-computer interface [R-BCI] that operates continuously during free behavior and generates activity-dependent stimulation of the brain or muscles. This so-called “Neurochip” consists of battery-powered electronics connected to electrodes that record the activity of motor cortex cells and/or muscles. The neural activity is processed by a programmable computer chip and can be converted in real-time to activity-contingent electrical stimuli delivered to nervous system sites or muscles (Mavoori et al, J. Neurosci. Meth. 148: 71, 2005). A promising application is to bridge impaired biological connections, as demonstrated for cortically controlled electrical stimulation of paralyzed forearm muscles. A recent study (Moritz et al, Nature 456: 639 – 642, 2008) showed that learning volitional control of neural activity that directly activates muscles is a promising alternative to the traditional decoding of neural populations for BCI control. A second application of the R-BCI is to produce Hebbian synaptic plasticity through spike-triggered stimulation, which can strengthen physiological connections (Jackson et al, Nature, 444: 56-60, 2006). Recent work has shown that similar plastic changes can be produced by EMG-triggered cortical stimulation and in the strength of corticospinal connections by cortically triggered intraspinal stimulation. The novel R-BCI paradigm has numerous potential applications, depending on the input signals, the computed transform and the output targets.

Support: NIH, Christopher and Dana Reeve Foundation, LSDF, ITHS, AHA, JST.

Factors Affecting Cursor Control by Small Ensembles of Motor Cortex Neurons

Marc H. Schieber, MD, PhD^2,3 and Andrew Law^1,2
1Department of Biomedical Engineering, University of Rochester, Rochester, NY
²Department of Neurobiology and Anatomy, University of Rochester, Rochester, NY
³Department of Neurology, University of Rochester, Rochester, NY

Recent studies of closed-loop control have shown that many primary motor cortex (M1) neurons rapidly alter their movement-related tuning as a normal subject learns to control a brain-computer interface (BCI). While many M1 neurons are known to be dissociable from muscle activity in experimental paradigms that dissociate a) visual targets from movement, b) kinematics from kinetics, or even c) neuron from muscle activity, little is known about which M1 neurons are most adaptable for control of a closed-loop BCI. We therefore are extending earlier studies of direct operant conditioning of single-neuron activity to identify factors that affect the ability of small ensembles of M1 neurons to control a closed-loop BCI. In a single session, the firing rates of 1 to 4 arbitrarily selected M1 neurons are combined in a linear transfer function that determines the one-dimensional motion of a cursor, and the non-human primate subject then uses the cursor to acquire targets of gradually diminishing size. Factors under study include: i) the physical distance between neurons, ii) the similarity of the neurons’ preferred directions during a center-out paradigm performed with the native limb, iii) the neurons’ discharge correlation during center-out performance, iv) the phasic versus tonic nature of the neurons’ discharge, and v) whether the neurons are fast- or regular-spiking.

Brain Flexibility In Neuroprosthetics

Dawn Taylor, PhD^1,2,3, Amar Marathe, MS^1,2,3, Stephen Foldes, MS^2,3 and Harrison Kalodimos^1,2,3
1The Cleveland Clinic Dept. of Neurosciences, Cleveland, OH
²Cleveland FES Center of Excellence, Cleveland VA Medical Center, OH
³Case Western Reserve University Dept. of Biomedical Engineering, Cleveland, OH

People commonly use one movement of the body to generate a different movement in a device or tool. For example, we adjust the position of a joystick to control the velocity of an object in a video game. We control the speed of our cars by adjusting the position of our foot over the range of the gas pedal. Our natural ability to make transformations during tool use suggests we may be able to take advantage of analogous transformations in brain-controlled devices. A wide range of movement-related signals can be decoded from the brain (e.g. velocity, position, reach goal, joint angles, muscle-activation levels). How do we make use of the best motor-related signals to control the device actions we want to control? Our lab is exploring the range of brain-to-device transformations people can learn to make, and how inaccuracies in decoding the intended movements can impact device control under different kinds of transformations. In many cases, neuroprosthetic control can be significantly improved by decoding one aspect of movement and applying it to the control of a different motor action. Visual feedback from the device enables individuals to learn new transformations through practice. Once a person learns to control a device, visual feedback continues to be used to correct for trajectory deviations in real time. However, while we easily learn many new motor transformations, we appear less able to learn certain error correction strategies necessary for adapting to the types of errors generated by

SESSION II: New Developments in Bioengineering and Materials

Toward High-Performance Cortically-Controlled Continuous Prostheses

Krishna V. Shenoy, PhD¹, Vikash Gilja, PhD¹, Paul Nuyujukian¹, Cindy A. Chestek, PhD¹, John P. Cunningham, PhD^1,2, Byron M. Yu, PhD^1,3,4, Joline Fan¹ and Stephen I. Ryu, MD^1,5
1Stanford University, Stanford, CA
²Cambridge University, Cambridge, UK
³University College London, London, UK
⁴Carnegie Mellon University, Pittsburgh, PA
⁵Palo Alto Medical Foundation, Palo Alto, CA

In recent years, cortically-controlled prostheses – which translate action potentials from neurons in the motor cortices into control signals for guiding computer cursors and robotic arms – have demonstrated considerable potential through a series of proof-of-concept laboratory animal experiments as well as an initial human clinical trial. While encouraging, several potential barriers remain which, if left unaddressed, may hamper the translation of these systems into widespread clinical use. First, an array of electrodes implanted in cortex (of rhesus monkeys, or humans) typically provides action potentials from highly-distinguishable single neurons for just a year or two, and even while working well the recorded signals grow and shrink on both slow (e.g., diurnal cycle) and fast (e.g., head movement) timescales. Second, the speed and accuracy of cortically-controlled computer cursor and robotic arm movements is much slower and less accurate than that of the natural arm. Third, the performance of cortically-controlled prosthetic devices has yet to achieve a level of robustness – the capability of running for hours straight, working seamlessly across days, and working across multiple behavioral contexts without human technical intervention – necessary for widespread clinical use. To address these three potential barriers, we conducted experiments with two rhesus monkeys with a 96-electrode array implanted in PMd/M1 and found that (1) threshold crossing detection provides high signal quality for many years, and with low fluctuation, (2) a continuous-decode algorithm redesigned (using a feedback control perspective) can provide cortical-cursor control on par with typical computer-mouse control, and (3) multi-hour, multi-day, and multi-context operation is readily possible.

An Anatomical Prosthetic Hand for Understanding Neural Control

Yoky Matsuoka, PhD
University of Washington

The Anatomically Correct Testbed (ACT) Robotic Hand was built as a tool to simulate dexterous physical world interaction based on direct or simulated neural signals. In return, the behavior expressed by the ACT hand uncovers mechanisms and neural control salient features in human hands that allow robust, versatile, and dexterous movements as well as rich object/world exploration. This talk describes this unique tool as well as things we learned so far, things we are learning now, and ways other researchers can contribute to future discoveries with the ACT hand.

Controlling Brain Circuits with Light

Ed Boyden, PhD
MIT Media Lab, Brain and Cog. Sci., Biological Engineering, and McGovern Institute, MIT, Cambridge, MA

The ability to enter information precisely into specific cell types and pathways within the brain would support the creation of neural prosthetics of potentially great power and flexibility. Over the last several years our group has developed a suite of genetically-encoded reagents that, when expressed in specific neurons in the brain, enable them to be activated or silenced in response to differently-colored pulses of light. These ‘optogenetic’ reagents are in use by hundreds of groups around the world, and we have demonstrated the safe and effective use of these reagents in the mammalian brain for controlling neural circuit dynamics downstream of given cell types. In order to enable these tools to be used for the systematic engineering of neural computations, behaviors, and candidate treatments for brain disorders, we have developed hardware to enable neural circuits to be perturbed in a three-dimensional fashion, and for measuring the brainwide neural dynamics impact of perturbing a specific cell class embedded within a neural circuit. We explore how these tools can be used to precisely alter the dynamics of neural circuits that mediate emotion, sensation, and movement, thus revealing principles that could be useful in the treatment of neurological and psychiatric disorders, and potentially in human augmentation.

Advanced Implantable Microelectrode Technologies for High-Fidelity, Multi-Modal Neural Interfaces

Daryl Kipke, PhD
University of Michigan

Technological advances in implantable neural interfaces are providing increasingly more powerful ‘toolkits’ of designs, materials, components, and integrated devices for establishing high-fidelity chronic neural interfaces for recording, stimulation, neurochemical sensing, and targeted drug delivery. Beyond progressive improvements in MEMS-based neural probe technologies, our group is developing new types of implantable microelectrodes using advanced nanostructured materials to obtain high-quality chronic neural recordings using structures that have a significantly reduced footprint compared to conventional microelectrode arrays. We are also developing site-selective, thin-film electrode coatings to make multi-modal microelectrode arrays for concurrent neural recording and neurochemical sensing at the microscale with high temporal resolution. These advanced technologies are extending the capabilities for precise, reliable, and high-fidelity neural interfacing in the brain. This research is supported by the U.S. National Institutes for Health and DARPA.

Session III: EcoG, EEG, and Less Invasive Approaches

Brain-Computer Interfaces: Theory vs. Reality

Jonathan R. Wolpaw, MD
Wadsworth Center, New York State Department of Health, Albany, NY

Brain-computer interfaces (BCIs) may provide valuable new communication and control
options for people with severe motor disabilities. Much BCI research has been based on
four assumptions: (1) that intended actions are fully represented in the cerebral cortex;
(2) that neuronal action potentials provide the best picture of them; (3) that, therefore,
the best BCI is one that records action potentials and decodes them; and (4) that
continuing mutual adaptation by the BCI user and the BCI system is not important. It is
increasingly clear that none of these assumptions is defensible. Intended actions are
the products of many areas, from the cortex to the spinal cord, and the contributions of
the different areas change continually as the CNS adapts to improve performance. BCIs
need to track and guide these adaptations if they are to achieve and maintain good
performance. Furthermore, it is not yet clear which categories of brain signals will prove
most effective for which BCI applications. In human studies to date, low-resolution EEGbased
BCIs and high-resolution cortical neuron-based BCIs perform similarly. In sum,
BCIs allow their users to develop new skills in which the users achieve their intentions
through brain signals rather than muscles. Thus, the primary task in BCI development is
to determine which brain signals users can best control, to maximize that control, and to
translate it accurately and reliably into actions that accomplish the users’ intentions. The
most difficult aspect of this task is probably not the realization of many degrees of
freedom, but rather the achievement of highly reliable performance. Much better
reliability is essential if BCIs are to advance from laboratory demonstrations to systems
of significant practical value in daily life.

Electrocorticographic Brain Computer Interfaces

Daniel Moran, PhD
Washington University, St. Louis, MO

Brain computer interface (BCI) technology has classically focused on two signal acquisition modalities for control: multi, single-unit activity (MSU) and electroencephalography (EEG). While MSU activity provides arguably the best multi-dimensional signal for BCI control, obtaining long-term stability of single unit recordings has proven difficult due to glial encapsulation issues. EEG, on the other hand, is a non-invasive technique where relatively large electrodes are placed on the surface of the scalp to record ensemble activity emanating from the underlying cortex. Given the large separation between the cortical surface and the recording electrodes as well as the inhomogeneous conductivity of the dura, skull, and skin; a rather large area of cortex needs to be synchronously active to be “electrically visible” on the scalp (~6 cm^2). Training such large cortical networks for BCI control of a few degrees-of-freedom can take month(s) to learn. On the other hand, electrocorticography taken from the surface of the brain (ECoG) are typically generated from much smaller neural ensembles (1-2 mm^2). Our recent results in non-human primates shows that epidural ECoG spectral power in the 60-200 Hz range is well correlated with ensemble single unit activity. Over a period of one week, the subjects learned to accurately control a 2D computer cursor through neural adaptation of microECoG signals over “cortical control columns” having diameters on a the order of a few mm. These results suggest that the mildly-invasive epidural microECoG is a pragmatic and possibly optimal modality for controlling neuroprosthetic devices.

**Additional abstracts coming soon.

Conference Location:

The Given Institute
100 East Francis Street
Aspen, CO 81611
Local Map

Directions

Public Transportation (Recommended)

RFTA (Roaring Fork Transportation Authority):
All points in Aspen can be reached using the RFTA bus system. Within the City of Aspen and many routes running between Aspen and Snowmass Village are free. For more information, click here.

Taxi Services

High Mountain Taxi: 970.925.8294

Mountain Valley Dial-A-Ride: 970.920.9999

Transportation to and from Aspen/Pitkin Airport

Most hotels in the city of Aspen provide complimentary shuttle service to and from Aspen/Pitkin Airport. Please inquire within at your individual hotel of choice upon book accommodations.

Parking

The City of Aspen implemented a transportation plan and parking management system designed to decrease congestion, improve air quality and preserve our small town character. Aspen's parking department monitors and enforces on-street parking programs, providing courteous and responsive customer service and information on all your parking options. Please click here for more information.

Suggested Hotel Accommodations in Aspen

(within a few blocks walking distance of the The Given Institute)

Hotel Aspen
110 W. Main Street
Aspen, CO 81611
Phone: 800.527.7369
Distance to Given Institute: .2 miles
Please click here for a walking map

Molly Gibson Lodge
101 W. Main Street
Aspen, Colorado 81611
Phone: 888.649.5982
Distance to Given Institute: .2 miles
Please click here for a walking map

The Annabelle Inn
232 W. Main Street
Aspen, CO 81611
Phone: 970.925.3822
Distance to Given Institute: .3 miles
Please click here for a walking map

The Limelight Lodge
355 S. Monarch Street
Aspen, CO 81611
Phone: 800.433.0832
Distance to Given Institute: .4 miles
Please click here for a walking map

*Be sure to mention that you are meeting at the Given Institute when booking your room*

For more information about these and other hotels in the Aspen area, please click here.

For general information about Aspen, please click here.

Special Needs and Additional Information

For any additional information and for special needs, including child/family care resources available to conference attendees, please e-mail Deanna Vollmer or call Deanna Vollmer 212.298.8611.