On Wed, 15 Apr 1992, Robert James Johnston wrote:
 
> Folks:
>
> I have heard of a group at UC Davis that is using FFTs for
> quantitative image analysis.  Applications include blood typing and
> virus identification.  One of my professors had some of their
> literature but misplaced it (you know how professors are...)
> Anybody out there know who these guys are?  I'm especially
> interested from the engineering point of view.
>
> Robert 'Jamie' Johnston
> [log in to unmask]
 
I don't know specifically who is looking at the FFT issue at UCD but
people have tried fft's for a long time in attempts to automate the
classifications that humans do quite easily. The basic premise is that the
morphology of the cell nucleus is often an indicator of cells gone wild.
One might expect 2-D fft's to be a reasonable algorithm for deciphering
the chaos in a neoplastic cell. Unfortunately, many people have spent the
better parts of their lives trying to develop automated cell cytometers
that actually work diagnostically. Probably the most successful individual
in this field is Peter Bartels (I think he is now at Arizona or Arizona
State). In terms of an actual device, one of the more successful groups is
NeoPath (Belllevue, WA). They have created a fully automated microscope
system with proprietary algorithms for screening "normal" PAP smears. The
concept here is that it is alot easier to screen out normals than it is to
give a definitive diagnosis on abnormal samples. By screening out the
normals, one can drastically reduce the number of slides that need to be
reviewed by a cytotechnologist and can then increase productivity/reduce
costs.
 
Paul Goodwin
Manager, Image Analysis
Fred Hutchinson Cancer Research Center
[log in to unmask]