Thanks for all the references. My overall impression by scanning
medical applications is that they tend to fall into the category of
"decision support". I once was particularly interested in this end
of medical informatics. I studied and did research work with neural
networks during college around '92-93. In '94, during medical school,
I feebly attempted to construct a system to interpret ploidy data from
a flow cytometer. Anyway, after that debacle, I let the whole medical
informatics interest lie fallow until coming across the electronic
medical records issue (in particular, this mailing list), as inspired by
my frustrations with (mostly paper) medical records during residency.
So...my initial posting on this topic was prompted by the statements
by the Columbia folks on the labor intensive and awkward process
of adding new clinical terms and concepts to a semantic system.
I think that probably a neural network approach would not be rigid
enough and would be too processor intensive to support computer-to-
computer interactions (though I am wishful for such an approach).
However, on a slightly different subject, it seems that the human-to-
computer interface might be greatly aided in regard to automatic
delivery of data to clinicians from disparate clinical databases,
through the analysis by neural networks or similar techniques.
So whaddya think?
Sincerely,
Robert
application/ms-tnef
