On Fri, 05 Jan 2001 10:44:13 +0000, "P.G.Hamer"
<[EMAIL PROTECTED]> wrote:
> Rich Ulrich wrote:
>
> > Computers do better than experts in making medical
> > diagnoses when the correct answer has to be from a narrow set.
>
> I think that some of the early systems also were better than humans
> at identifying the possibility of unusual diagnoses. AFAIR it took the
> humans to reach a final diagnosis, but the computer could raise options
> that the human might not have thought of. [Sort of reassuring if
> you have a medical emergency in the small hours, with only an
> overworked/underslept junior doctor to perform the diagnosis.]
>
> Again going back a few years, tropical diseases might not even be
> considered as a diagnosis for UK residents. Which might be a problem
> if you had just been to the tropics.
- I think I remember, one case was an example of an early "victory"
for a computer program tested against doctors.
>
> Computers are good at not forgetting all the possibilities.
- This morning, I asked a doctor, "Whatever happened to computer
diagnosing? It used to look so promising?" Then he spent a few
minutes telling me about it - "It's being used everywhere."
Computers give reports on what seems interesting in EEGs, or in
X-rays. That's one level of help. At the higher level, I remembered
that there had been questions - 15 years ago - of using "computer
diagnoses" because of legal liability. And, I thought, keeping
up-to-date had to be problematic.
It looks like that has been finessed by keeping the computer in the
background as another reference tool. My friend says that at
Mass.-General, every department has computers that are set up for
suggesting, testing and comparing differential diagnoses. However,
the doctor always has the final word. I had not been aware of all
that.
--
Rich Ulrich, [EMAIL PROTECTED]
http://www.pitt.edu/~wpilib/index.html
=================================================================
Instructions for joining and leaving this list and remarks about
the problem of INAPPROPRIATE MESSAGES are available at
http://jse.stat.ncsu.edu/
=================================================================
