91-04/Neural.info
From: "Marc W. Cygnus" <cygnus@cis.udel.edu>
Subject: Neural Interfacing (some current info)
Date: 29 Apr 91 04:14:32 GMT
Organization: UDel Artificial Life Group



Over the past week or so I've seen a few articles speculating on the progress
of the state-of-the-art in neural interfacing; following is some information
on "current" (mid-1990) research which I hope will be useful to anyone
interested in this field.  (I certainly am... you could call it a burning
obsession of sorts :-)  References are at the end of the posting.

In the May 1990 issue of _Science_, Science intern Sarah Williams reports on
a neural interface device developed at Stanford University [1].  The report
contains information presented by Gregory Kovacs at a 1990 plastic surgeons'
meeting held in May in Washington, D.C..

Apparently, she slightly misrepresented a few details in the report, because
in the June issue there's a letter from Kovacs in which he "[clarifies]
some statements made in... [the] article..." [2].  (nothing major, just
details)

Here is a summary of the information in the article, corrected where
necessary by drawing from Kovacs' letter:

  Gregory Kovacs (Stanford University), along with Joseph Rosen (Stanford),
  Bernard Widrow (Stanford), and Chris Storment (Dept. of Veterans Affairs)
  have tested a neural interface chip which allowed recording of action
  potentials from individual neurons in their experimental setup.
  
  The chip was a little slice of silicon onto which a square array of 1,024
  iridium microelectrodes were "stenciled."  Then, a "high-performance plasma
  etching process" [2] was employed to drill tiny holes through each pad and
  through the chip, after which the entire chip was coated with silicon
  nitride.

  In their experimental setup, they implanted the chip in a rat's leg by
  severing a nerve (presumably a "well-known" peripheral nerve), inserting
  the chip in the cleft, and allowing the nerve to regenerate; during the
  regeneration, individual nerve cells grew through the holes in the chip,
  thereby providing a microelectronic link to each axon's activity.

  Kovacs says in his letter,
      
        We make no claim to have been able to stimulate "individual neurons."
      While this may be possible with our device, our initial experiments
      were not designed to test this.  In the pilot study, we demonstrated
      recording from, and stimulation of, peripheral nerves.  We believe
      that we were able to _record_ action potentials from individual
      neurons.  However, there is a big difference between stimulating and
      recording.  Current work is focused on determining how selective the
      devices are in both of these modes. [2]

  The last paragraph of his letter is perhaps more important to those of us
  wishing to understand the state of progress in this field.  He says,

        Attempts to fabricate and use such neural interfaces are not new.
      Since the early 1960s experiments have been conducted along those lines,
      but only recently have fabrication techniques been developed that allow
      devices to survive in the body for extended periods.  Interfacing to
      the nervous system will undoubtedly be done sooner or later, with or
      without this project.  The only claim we make is that we are doing our
      best to achieve this goal. [2]


Another big advance related to the problem of direct neural interfacing came
about fairly recently but I cannot remember my source.  If anyone knows of
the research I'm describing in the following sentences, please email me!  If
not, I'm sure I can dig up the references given a little time.  So, for those
of you who don't do this automatically for missing references, please *take-
what-i-say-with-a-grain-of-salt*, because this is strictly from memory.

Anyway, the work has to do with the fact that cells of the CNS aren't happy
regenerating (one of the reasons spinal cord injuries are so traumatic).
The reason CNS cells don't regenerate has apparently been either discovered
or more precisely defined: it's not that they don't regenerate, it's that
the body secretes a growth-suppression factor which keeps them from regen-
erating.  A research group has found an anti growth-suppression factor which
either suppresses or negates the effects of the natural factor; they have
reported regeneration success in an experiment where they severed the spinal
cord of a rat to which the anti growth-suppression factor was administered.
I want to say the experiment involved actually _removing_ a small (>1mm or
less) section of nerve so that the ends weren't touching, but I'm not really
sure about that.

The stuff above, in conjunction with the Stanford experiments, is tremendously
exciting, at least for me.  Of course, there exist complicating questions and
problems beyond those associated with simply "tapping into" a single neuron.
Is that really what should be done?  A big problem there is the fact that
science has yet to really make a dent in the neural connectivity problem.
It's one thing to have action potential information for every single axon in
a bundle, but it's an entirely different thing to assimilate that information
into something meaningful.  Much more often in research it's population
potentials (intercellular potentials resulting from the combined microcurrents
through a small population of neurons) which are correlated to events in
the physical world.  Then again, the field of neural networks (in the silicon
context, not the biological :-) might likely hold a solution to the inter-
pretation problem.

I fancy the applications to VR interfaces, if (*when*!) that time comes, will
appear long after rehabilitative applications are perfected, but progress is
after all progress!

                                        -marcus-

ps: quotes taken without permission from issues of _Science_. I looked for
copyright restrictions in the magazine but found none relating to information
redistribution.

--------------------------
1. "Tapping into Nerve Conversations" (Research News), _Science_ 248,
        p. 555 (4 May 1990).  There's a good uphoto of the earlier
        64-electrode prototype chip.
2. "Neural Interfacing" (Letters), _Science_ 248, pps. 1280-1281
        (15 June 1990).


-- 
-----------------------------------------------------------------------------
"Opinions expressed above are not necessarily those of anyone in particular."
 UDel Artificial Life Group (Graphics Support)  |  INET: cygnus@cis.udel.edu
 114a Wolf Hall (Irisville) (302) 451-6993      | CompSciLab: (302) 451-6339

