🚀 go-pugleaf

What follows is a brief description of the work done at
the Neurodynamics Lab at SUNY/Health Science Center Brooklyn.

Our lab measures evoked-potentials: human brain activity
in response to stimuli, by placing an array of electrodes
on the subject's scalp and showing a variety of images on
a video monitor or providing auditory stimuli through
earphone.  The stimuli are typically 1.5 to 3 seconds apart
and may be geometrical figures, words, or pictures of faces,
(visual) or simple tones.  A typical experiment involves
several hundred stimuli, and the subject is asked to provide
a response, by pushing a button, to some or all of the 
stimuli.
	The peak response to a stimulus will involve a
voltage change of about 20-40 microvolts.  The current from
the electrodes is fed to a 10e5 amplifier and to a 12-bit
a/d converter.  Samples are 256/sec, generally for the
duration of the experiment, although sometimes there are
gaps.  The data is digitally filtered online and shown to
the operator, so that bad electrodes can be detected and
fixed.  All data is saved, and is also summed into the small
number (2-8) of cases into which the stimuli are divided.
The amplitude and latency of maxima and minima of typical
waveforms are the data of significance, but topographic
maps of current density are also constructed, and some
effort has been made to identify dipole sources inside the
skull.
	This project is being replicated in part in a number
of labs across the country with identical equipment and
software (written by me) underwritten by the NIAAA as part
of a study of the genetics of alcoholism.  It will result in
a vary large amount of data on human evoked potentials.
The rationale is that earlier studies at this lab have
shown that there are characteristic differences in the
evoked potential data between subjects at risk for alcoholism
and those not.  (These differences are not an effect of
alcoholism.)  In addition, evoked potential patterns are 
highly heritable.  So we hope to use this data to tell the
geneticists at whom to look for significant differences.
The problem that this raises is how to classify individuals
based on a large amount of data on each person that we will
have.  Any suggestions or pointers to literature?
	The data acquisition and stimulus presentation is done
on a Concurrent 5550 running RTU (basically ATT Unix with a
number of real-time extensions).  This is an old Multibus
68020 machine with a lot of extra goodies built in:  data
acquistion hardware, vector processor, etc. 
The one hardware feature of
this machine which is essential is that there is a separately
programmable graphics processor, so that as the main cpu
handles data acquisition, the graphics processor independently
runs a program which shows the stimuli to the subject. (Most
graphics processors run by executing a sequence of commands
from the main cpu.)  Most of the analytical work is done on
Sun workstations and servers.

David B. Chorlian Neurodynamics Lab  SUNY/HSCB