(Excerpted from my NSF Graduate Fellowship application)
Bidirectional Neuroprostheses
Introduction
As a neuroengineer, I want to design prosthetic devices that interface with the
brain. While the first generation of neuroprostheses have focused on replacing
lost sensory functions on the periphery of the nervous system, the next
generation will engage in bidirectional communication with intrinsic neural
circuits. This new paradigm demands a more detailed understanding of the
underlying networks of neurons, as well as sophisticated microelectronic design
techniques to create small, powerful, and efficient signal-processing devices.
Success in this field will require highly interdisciplinary work and expertise
in both electrical engineering and neuroscience. Here I propose a research plan
that begins to explore one application of bidirectional neuroprostheses.
Motivation and Background Information
Bidirectional neuroprostheses are a novel class of devices capable of receiving
input from, and sending output to, neurons in the central nervous system (CNS).
In the case of trauma to the CNS, a future neuroprosthesis might communicate
with intact cells and emulate the properties of the damaged region, thereby
restoring function. Traditionally, sophisticated signal-processing algorithms
are used to interpret neural activity, but these methods require significant
computational power and impose a rigid mathematical structure to explain the
dynamic operation of thousands of neurons. A potentially more appropriate
solution for neuroprosthetic applications is to generate computational models
that mimic the operation of a network of cells, and then reconstruct the neural
circuits in silico using neuromorphic very-large scale integration (VLSI) design
techniques [8].
Neuromorphic circuits resemble their biological counterparts in both structure
and function. By utilizing the intrinsic properties of silicon transistors in
low-power operation, small circuit elements can exhibit current-voltage
relationships similar to those observed in real neurons. Thus, the complex
calculations performed by individual cells come “for free” in this paradigm,
enabling a designer to concentrate on the emergent signal-processing
capabilities of the network. In addition, neuromorphic circuits can operate in
parallel and function in real-time. All of these properties make neuromorphic
VLSI technology an ideal candidate for implementing implantable neuroprostheses.
Research Plan
Goals
My long-term goal is to create a prosthetic device that will re-enable
locomotion in paraplegic individuals. It is well known that the spinal cord (SC)
of vertebrates contains local networks of neurons that are hardwired to execute
various motor “programs”. These circuits, called central pattern generators (CPG),
facilitate the sequential activation of muscle groups required to produce
complex behaviors such as breathing, chewing, locomoting, and other rhythmic
movements [1]. Spinal injury disrupts the descending pathways used by the brain
to activate CPGs, but leaves most local circuits intact. Therefore, it may be
possible to restore motor function after spinal injury by designing a
neuroprosthesis capable of interfacing with and controlling intrinsic CPG
networks [4]. I believe that my educational background combined with my
advisors’ expertise in these areas makes me uniquely suited to conduct this
research.
As a graduate student, I will work with an animal model of spinal injury.
Specifically, I plan to investigate CPG networks in the lamprey, a well-studied
primitive vertebrate. The lamprey CPG consists of a chain of approximately 100
coupled oscillators called unit pattern generators (uPG), each located within a
single segment of the spinal cord. In an intact specimen, all of the oscillators
work in tandem to produce a traveling wave of activity that alternately excites
the motor neurons on opposite sides of the body, producing a sinusoidal pattern
of muscle contraction that propels the animal through the water [3]. In the case
of spinal injury, one or more spinal segments caudal to the lesion site become
isolated from the rest of the SC and no longer receive activation or
synchronization signals. To restore function in this situation, it is necessary
to first initiate and then coordinate the oscillations between the regions
rostral and caudal to the location of trauma. Because naturally-occurring uPG
circuits perform these duties efficiently and effectively, I intend to employ a
neuromorphic microchip to execute these tasks.
Specific Aims
My proposed graduate research would make significant contributions toward
understanding and controlling spinal CPG networks. As I work to create a
neuroprosthesis capable of restoring motor function after chronic spinal injury,
I have three short-term aims. First, I hope to improve on existing neural models
of CPGs [5] by incorporating biologically accurate feedback into the cellular
network. This will be an essential component of my work, as real-time signals
from recording electrodes will be used extensively to modulate the activity of
the neuroprosthesis, improving the coordination between the silicon hardware and
biological “wetware”.
Second, I intend to determine the appropriate parameters for effective
electrical stimulation of spinal CPG networks so that I can generate a
phase-response curve (PRC) for the lamprey SC. A PRC is a mathematical
description of the effects of perturbing an oscillating circuit at various times
during its activity cycle [7]. Once I am able to characterize the lamprey CPG in
this way, I can employ a number of analytical tools to evaluate potential
control strategies.
Third, I wish to create a neuroprosthetic device that enables the functional
coupling of two halves of a severed lamprey SC in an isolated (in vitro) spinal
cord preparation. This neuromorphic microchip will translate the sophisticated
CPG model described in my first aim into compact, low-power, silicon VLSI
hardware. My circuits will improve on previous designs of neuromorphic CPGs [6,
2] by including adaptive capabilities, accepting biological inputs, and
providing appropriate outputs for directly stimulating nervous tissue.
Discussion
Much of my work on the spinal injury model in lampreys is likely to be
applicable to understanding motor control and restoring motor function in
humans. For example, knowledge about the fundamental organization of CPG
circuits in primitive vertebrates will facilitate the identification of their
counterparts in higher organisms. Additionally, by mathematically characterizing
the effects of stimulation on the lamprey CPG, we will become better equipped to
interact with such networks and emulate their function. Integrating sensory
feedback into the CPG control mechanism is also important for future
experiments: when we work with freely-moving animals in unpredictable external
conditions, our microchips must be able to detect deviations from the expected
behavior and respond rapidly. Most importantly, the experience I gain through
this project should perfectly prepare me for a career devoted to designing and
implementing bidirectional neuroprostheses. The outcome of this work will have
ramifications for both basic science and people suffering from spinal cord
injuries.
References
[1] F. Delcomyn. Neural basis of rhythmic behavior in animals. Science,
210(4469):492–498, 1980. {back}
[2] S. P. DeWeerth, G. N. Patel, M. F. Simoni, D. E. Schimmel, and R. L.
Calabrese. A VLSI architecture for modeling intersegmental coordination. In R.
Brown and A. Ishii, editors, 17th Conference on Advanced Research in VLSI, pages
182–200, Ann Arbor, Michigan, September 1997. IEEE Computer Society.
{back}
[3] S. Grillner, T. Deliagina, O. Ekeberg, A. El Manira, R. Hill, A. Lansner, G.
Orlovsky, and P. Wallen. Neural networks that coordinate locomotion and body
orientation in lamprey. Trends in Neurosciences, 18:270–279, 1995.
{back}
[4] R. Jung, E. J. Brauer, and J. J. Abbas. Real-time interaction between a
neuromorphic electronic circuit and the spinal cord. IEEE Transactions on Neural
Systems and Rehabilitation Engineering, 9(3):319–326, 2001.
{back}
[5] R. Jung, T. Kiemel, and A. H. Cohen. Dynamic behavior of a neural network
model of locomotor control in the lamprey. Journal of Neurophysiology,
75(3):1074–1086, 1996. {back}
[6] M. A. Lewis, M. J. Hartmann, R. Etienne-Cummings, and A. H. Cohen. Control
of a robot leg with an adaptive aVLSI CPG chip. Neurocomputing, 38-40:1409–1421,
2001. {back}
[7] A. D. McClellan and W. Jang. Mechanosensory inputs to the central pattern
generators for locomotion in the lamprey spinal cord: Resetting, entrainment,
and computer modeling. Journal of Neurophysiology, 70(6):2442–2454, 1993.
{back}
[8] C. Mead. Analog VLSI and Neural Systems. Addison-Wesley, Reading,
Massachusetts, 1989. {back}