|
His first job was at Rush-Presbyterian-St. Lukes Hospital, a teaching
hospital. Stereotypical of first jobs, he stayed for exactly
one year, from September of 1975 to September of 1976. During that short
time, however, he worked on many different, cutting edge systems (thanks
Lauren !).
At the time Rush was abuzz with technology projects (they may still
be for all I know). Their MIS group was at the cutting edge of Hospital
Information Systems development. The hospital even operated back then
what today would be called an HMO.
Pacemakers, Big As Cigarette Packs
Matt developed a production application for tracking pacemaker patients.
The software ran on a General Automation SPC-16, a knockoff of the IBM
1800 16-bit minicomputer; the software was written in a variant of Fortran
IV (or was it V? :-)
CAT Scanners (but he's allergic to cats!?)
He got to play with a CT scanner, which
at the time was one of three in the US; there weren't that many more
in existance anywhere else in the world back then. They were (all!)
manufactured by EMI (i.e. EMI had 100% of the fledgling market).
EMI, however, fell asleep at the wheel and GE ate its lunch; to this
day GE is the leader in CT scanners and EMI is back at publishing music
(have they switched to CDs, yet?). His “job” was to help write
software to use the output of the CT scanner to calculate the volume
of ventricles in the brain.
The scanner's output was a digital “slice” of, in this case,
the brain. Each voxel had a dynamic range of
less than 26. That meant that the resulting image was, shall
we say, coarse?  .
The goal was to “stack up” several of these slices to create
a 3D representation of the brain and from that to calculate the volume
of ventricles in the brain. Of course, there were a couple of obstacles
that had to be worked out,
- it was not feasible to use a live brain to develop the program
since it would be awkward to scan a human being repeatedly, and
- to debug the program and the algorithms meant that they needed
to know the real position and size of ventricles a priori.
The workaround to this inconvenience was to use a real skull, a few
baloons, and jell-o. The skull came from an anatomy class. The brain
cavity was filled with jell-o plus some extra stuff to make it about
the consistency of a live brain. In the brain, er, jell-o they quickly
had to place a balloon with inflated to a known volume and shape and
place it in a known location (e.g. where a real ventricle would
appear in a diseased brain). Once the jell-o hardened, the skull was
wrapped in towels and run through the scanner.
All I can say is that it was fun!
The system “controller” really was a Data General (DG) Nova,
8-bit minicomputer (auto-decrement and auto-increment memory
locations, yo!). Later those were replaced with Eclipse, 16-bit minicomputers
(with this computer, DG hoped to “eclipse” DEC, er …
I mean, Digital — of course, neither Data General, nor DEC needed
any help to eclipse themselves).
The Xeromamothing
The other project I got to work on was a breast cancer diagnostic
system (well, OK, a prototype). This was the brainchild of Lauren Ackerman,
my boss at the time. The system ran on a PDP-11/40 clone, manufactured
by a California company called Caldata. This
thing was the only one of its kind at Rush (although a few more existed
at the nearby University of Illinois campuses) and it ran Unix! (version
five, though Matt immediately upgraded it to version six :-).
This Unix machine was the controller for a contraption that was supposed
to digitize Xeromamograms (i.e. X-rays of breats, printed on
paper, not film) via a (very) slow scan camera. The Xeromamograms sat
upright in a wooden tub. At one end of the tub there sat a camera and
at the other end there was a metal box, as big as a mamogram, attached
to (the guts of) a vacuum cleaner. A mechanical grappler moved (some
of the time) back and forth over the mamograms and it would come down
over the right one (maybe), grab it (perhaps), and pull it out (or rip
it out, as the case may be). It then moved it away from the camera to
the end of the tub where the mamogram was sucked flat onto the metal
grill by the vacuum cleaner running behind it. The camera would then
ever-so-slowly digitize the mamogram. The A/D (Analog-to-Digital) Converter
was in a little box, attached to the Caldata via a 16-bit parallel interface
(DR-11C?). Software in the clone then attempted to determine if there
was something that looked like a tumor present, and, if there was one,
to determine from its shape whether it was malignant or benign.
Etcetera, Etcetera.
In his spare time, Matt managed to program all sorts of devices: a
(big, big) calculator with a dixie (sic?) tube display; a (big) HP digital
calculator that could actually handle alphanumerics; an ecocardiogram
machine; the IBM 360 (what model?) in the data center; and, an HP 1000,
used in angiogram procedures.
After about a year at Rush, Matt heard of a Unix shop opening at the
University of Illinois Medical Center. Problem was, they were not going
to hire “mere programmers” until they hired a manager. So,
he applied for the manager position. And got it (and was as surprised
as anyone!).
Along with a very good friend, Ray Roth, and a couple of other folks,
they ran an installation consisting of a PDP-11/50 (32K of semiconductor
memory, 32K of core, and a floating point unit — hot stuff), a
PDP-11/34 (64K of semiconductor memory), and a PDP-8 (can you say “TECO?”).
This period lasted for over four years. They were a blast and an opportunity
to try all kinds of things out and work on very challenging projects.
We were even connected to the DARPA
network then (if you don't know that the DARPA network is the direct
predecessor of the Internet and all that’s wonderful on this earth
… where have you been? hiding under a rock? and you couldn't get
broadband installed in the rock?)
|
