|
A. Abstract
The ability to learn a person's intentions by "reading" his
or her brain activity has always been a fascinating theme in the realm
of science fiction. Yet, recent advances in brain research indicate
that this idea may not be so far-fetched after all. Moreover, while
the ability to control brain function at first sight appears frightening,
it may also turn out to be a blessing for mankind: A benign usage
of such knowledge has the potential to relieve various forms of disabilities.
Thus, one of the exciting venues of current research is the develo
pment of motor neural prostheses, which will ‘read’ brain activity
and use the output to control the movements of a paralyzed limb or
prosthesis. Examining state-of-the-art studies, we find that a crucial
problem with this approach is our poor ability to extract the relevant
information from the monitored brain activity. The activity provides
only partial and "noisy" information about the subject's
intentions. Moreover, it changes continuously, due to technical problems
such as unstable recordings, or due to the inherent adaptive nature
of the brain itself, which modifies its activity to the subject’s
experience. The METACOMP project is a collaborative effort of teams
of scientists with diverse skills in neuroscience coordinated by Prof.
Eilon Vaadia in Israel and Prof Ad Aertsen in Germany. Our goal is
to be able to construct a system that can efficiently reconstruct
desired movements, in a way that best represents the subject’s intention.
We expect our results to provide a major step toward our long-range
objective - to advance towards the usage of these methodologies for
improving motor function in amputees, patients who suffer spinal cord
injury, Parkinson disease or other motor disorders that leave the
motor parts of the brain intact. We emphasize that, we do not expect
implementation in human subjects within the first time frame (5 years)
of this project. Yet, the prospects for future application are good.
Our work is at its early phase but has already yielded preliminary
exciting results. Our first efforts in the last year were to optimize
extraction of relevant information from neural activity. We made first
steps in this direction in two ways. First, we improved existing computational
tools to extract information from the neuronal activity1,2 and second,
we found how one can “improve the brain activity” by specific learning3.
These results give more hopes to understand what the brain "wants"
to tell the moving limb.
B. Main lines of research
There are at four main directions of research that are essential for achieving
the ambitious goal of brain-driven robotic devices.
(1) Develop techniques for simultaneous recordings of large numbers
of neurons as reliably as possible for prolonged periods of time.
While various types of microelectrode arrays are under development
and in use, the problem of the recording quality is still not solved.
This work will not be done in Vaadia lab since it requires very expensive
methodologies. However, Dr. Vaadia is negotiating with high-tech engineering
companies and will be willing to test their products.
(2) Improve our understanding of the neural code. It is still unclear
how neuronal activity in motor cortex represents and controls ongoing
movements. While some hold that populations of neurons accurately
represent the time course of movement kinematics (direction and speed
of the limb) others maintain that the motor cortex encodes intrinsic
dynamic parameters (joint by joint, muscle by muscle activation).
This project will develop and examine additional approaches tonalysis,
information extraction and interpretation of neuronal recordings.
(3) Develop algorithms to transform neuronal activity to reconstructed
movements. This will serve as the "model" in figure 1. Various
algorithms are good candidates for interpreting neuronal activity
and prediction of movements. The interdisciplinary center for neural
computation is especially gifted with scientists who have already
contributed to the efforts to develop improved methodologies. (see
example -
DDT animation)
(4) Interface robotic devices with the biological system. Several groups
have been engaged in experiments to develop the most challenging class
of prostheses mentioned above - brain-driven prostheses. While this
research is still at its first steps, it is evident that this challenge can be
met in the future. Our group will use available robotic devices.
One such device is already in our laboratory. We plan to use the knowledge
we acquire to control this robot by brain activity that will be recorded from
awake monkeys.
C. Outlook
When the first phase of this research is completed (5 years) we plan to have a
complete system that allows for controlling a robotic arm in real
time by reading brain activity in monkeys. No invasive experiments
in humans are planned within the scope of this project. Instead, we
rely on psychophysics (behavioral experiments) in humans, electrophysiology
in primates, and the development of novel tools for data processing
and adaptive control. Clinical tests may, however, be scheduled after
the principle has been successfully demonstrated in primates, and
after the issue of long-term bio-compatibility of the direct interface
has experienced enough progress. Our experiments will provide detailed
and quantitative information on the prerequisites that ensure stable
and robust performance of the envisaged prosthetic devices. This knowledge
will be exploited to design interfaces that allow minimizing the invasive
actions taken in future applications for human patients. |
|
|