Multi-electrode and optical techniques for studying real neural nets
promise to reveal many complex, dynamic properties of neural ensembles that
were difficult or impossible to study with conventional microelectrodes.
Unfortunately, there is an information bottleneck in storing and analyzing
the huge amount of data generated by such multichannel techniques. The
trend must go towards more on-line, parallel analysis and data reduction
if we are to make headway. Special-purpose spike-sorting hardware and
software have been created, using conventional template-matching algorithms.
It seems to me that this sort of problem is ideally suited for artificial
neural networks. Yet I have found only one paper in which NNs have been used
to analyze data from real neurons, and it was off-line:
"Data processing for multi-channel optical recording: action potential
detection by neural network" by S. Yamada, H. Kage, M. Nakashima, S.
Shiono, and M. Maeda, J. Neurosci. Methods. (1992) 43: 23-33.
Does anyone know about others who have applied artificial neural nets to
studying living neural nets, especially in terms of spike analysis in
real-time? If so, please email me directly (ie, dont reply to this newsgroup)
at:
spotter at darwin.bio.uci.edu
Thank you. I will post a summary of any replies I receive.
Steve Potter
Psychobiology dept.
UC Irvine
Irvine, CA 92717
