Download Presentation Summary - CAINS Center for Autonomous Intelligent

Theodore W. Berger
AINS Conference / 2002
Affiliation: Professor of Biomedical Engineering and Neuroscience, Director of
the Center for Neural Engineering, University of Southern California, Los
Angeles, CA
Title:
Toward Replacement Parts for the Brain: Implantable Biomimetic
Electronics
Abstract:
Dr. Berger will present results of a collaborative effort by
neuroscientists and engineers at the University of Southern California to develop
the technology required for hardware implementation of neural network models to
be used as neural prostheses for the replacement of damaged or dysfunctional
brain tissue. Five components of this effort will be described, which include (i)
experimental study of cellular and molecular mechanisms, (ii) formulation of
biologically realistic models of network dynamics, (iii) hardware implementation of
the network models in analog VLSI, and (iv) hybrid neuron-silicon devices for
interfacing brain implants with existing neural tissue. Described as part of the
latter effort are recent successes in culturing living neurons directly onto siliconbased computer chips. This five-part approach is applied to developing a
prosthetic device for the hippocampus, a region of the brain responsible for the
formation of long-term memories, and that frequently is damaged as a result of
epilepsy, stroke, and Alzheimer's disease. The hippocampus is typical of most
neural systems found in the mammalian brain in that it is composed of several
populations of neurons, each with a distinct set of functional properties that
include high order nonlinear dynamics, and are interconnected through a variety
of feedforward and feedback circuits. Because of its special role in memory
formation, the strengths of connections between hippocampal neurons also are
subject to activity-dependent modification, i.e., the system properties can be
nonstationary. In the intact hippocampus, subpopulations of neurons are
involved in distributed representations of stimuli and/or behavior that can be
altered with changing experience or context.
Collectively, these system
characteristics present a unique set of challenges to developing a biologically
realistic model at the systems level which also provides a sufficient basis for a
partial replacement neural prosthetic. Through multidisciplinary efforts of
neuroscientists, engineers, and medical researchers, many of these challenges
can be overcome with present-day technologies, to the extent that the first
applications of silicon-based computational elements having direct
communication with brain tissue will be realized in the near future. In this regard,
novel “neuromorphic” silicon-based multi-site electrode arrays have been
fabricated and tested as neuron-silicon interfaces. The spatial distribution of
electrode sites is specifically designed to be consistent with the cytoarchitecture
of the hippocampus, and brings the uniform distribution of microchip contact pads
into the register with the non-uniform distribution of hippocampal neurons. Dr.
Berger also will demonstrate with examples how research efforts to develop such
a new generation of neural prostheses inevitably "spin-off" other technologies
that are highly useful for real-world applications, e.g., speech recognition and
other temporal pattern identification systems.