FILE NAME:  CANADA.TX            5/25/73

AND AMINTAIN.
	LABORATORY COMPUTERS

A COMPUTER SUCH AS THE IBM 1800 IS INSTALLED WITH THE  
EXPECTATION THAT IT WILL ULTIMATELY SERVE  
FOR MANY PROJECTS AND MINIMIZE ANY REQUIREMENTS
FOR OTHER SPECIAL DATA ACQUISITION HARDWARE. THE INITIAL
COST IS HEAVY...BUT WE EXPECT RICH REWARDS.

WE HAVE CREATED A COMPUTER SYSTEM TO WORK WITH MANY PROJECTS,
WE HAVE BEGUN RUNNING A VERY LONG OBSTACLE COURSE.
THEREFORE IT IS COSTLY. NOW WE HAVE TO FIND AS MANY PROJECTS
AS POSSIBLE TO PUT ON THE SYSTEM TO JUSTIFY THE COST.
THE TEMPTATION IS TO GET THOSE PROJECTS ON EVEN IF
SPECS OF THE PROJECT ARE NOT A GOOD FIT TO THE COMPUTER.
THE TENDANCY IS TO TAILOR THE OBJECTIVES TO FIT THE MEANS.
THE NEED TO JUSTIFY THAT ORIGINAL COST IS BEGINNING TO DISTORT
THE PLANNING PROCESS. THERE MAY BE AN ENEMY IN THE HOUSE.

THE LESSON HERE IS DON'T BE OVERSOLD ON YOUR FIRST COMPUTER SYSTEM.
USE IT AS A TOOL FOR DETERMINING THE STEPS REQUIRED FOR AN 
ADDITIONAL LEVEL OF CAPABILITY. GIVE YOUR PEOPLE AN OPPORTUNITY
TO GROW AND CONTRIBUTE TO SPECIFICATIONS FOR COMPUTER SYSTEMS.
YOUR FRIENDLY COMPUTER SALESMAN IS AN OUTSIDER, WEDDED TO
THOSE DESIGNS THAT SELL THE LARGEST NUMBER OF DIGITAL COMPONENTS.
BEWARE!!


	DATA REDUCTION AT THE SOURCE

WE WANT TO GET THE MOST COMPUTER POWER AT THE MINIMUM COST.
WE WANT TO SIMPLIFY THE PROGRAMMERS JOB SO HE 
CAN FINISH AND GET ON TO THE NEXT ASSIGNMENT. WE WANT TO PUT THE 
MAXIMMUM NUMBER OF INPUTS ON THE COMPUTER . BUT HOW!!

WELL,  IF YOU NEED FLOODS OF REDUNDANT ANALOG DATA TO 
PULL OUT A FEW ESSENTIAL VALUES THE OBVIOUS STEP IS PREPROCESS
AT THE SOURCE.

AS AN EXAMPLE LETS TAKE AN AUTOANALYZER OUTPUT. WE DECIDE THAT
THE CURVE REQUIRES AT LEAST 100 POINTS TO DESCRIBE IT WITH 
ACCEPTABLE FIDELITY. WE ARE GOING TO EXAMINE THIS CURVE FOR
 THE PEAK VALUE. DATA REDUCTION VIA PREPROCESSING 
MEANS THAT THE COMPUTER WILL SEE ONLY ONE DATA WORD  EVERY 60 SECONDS
INSTEAD OF 100.

NOW LETS LOOK AT A THE BLOOD PRESSURE CURVE OBTAINED FROM A CANNULATED 
ARTERY. IT HAS A STEEP UPSTROKE TO ITS SYSTOLIC VALUE AND A MORE
GENTLE NEGATIVE SLOPE TOWARD THE DIASTOLIC VALUE. WE DECIDE
THAT 150 EVENLY SPACED SAMPLES ARE NEEDED TO DESCRIBE THIS 
CURVE AND THIS DATA MUST BE COLLECTED AT A RATE OF 150 SAMPLES
PER SECOND. IF WE PREPROCESS FOR SYSTOLIC AND 
DIASTOLIC VALUES FOR A HEART RATE OF 60 BEATS PER MINUTE, COMPUTER 
INPUT DROPS TO 120 PER MINUTE INSTEAD OF 900 PER MINUTE.
A NEW BENEFIT RESULTS. THE DATA RATE IS SO LOW WE CAN PLAN
TO TRANSMIT TO A REMOTE COMPUTER VIA AN EXTENSION PHONE 
IN THE LABORATORY. ALSO,  WHEN THE DATA RATE IS LOW ENOUGH WE CAN 
LOCATE THE LABORATORY OR THE COMPUTER WHEREVER CONVENIENT.
ANOTHER BENEFIT DERIVES FROM THIS, YOU CAN INCLUDE 
THE USE OF A BACK UP COMPUTER AS A NORMAL PART OF THE DESIGN
OF YOUR INFORMATION SYSTEM.

WHAT ELSE...WELL IF OUR DATA REDUCTION IS SUFFICIENT WE CAN 
MULTIPLEX SOURCES AT THE LABORATORY AND TRANSMIT THROUGH
ONE LINE TO THE COMPUTER...AND IN FACT WE DO ALL OF THESE
AND IT WORKS.


	INTRODUCTION

LETS TALK A LTTLE ABOUT PEAKS:WHAT THEY REPRESENT AND WHERE WE USE THEM.
ANALOG SIGNALS ARE DIGITIZED HUNDREDS OR THOUSANDS OF TIMES IN
THEIR TIME COURSE,BUT IN MANY APPLICATIONS
THE PEAK AMPLITUDE AND PERHAPS ITS LOCATION ARE THE 
ONLY DATA ESSENTIAL FOR FURTHER PROCESSING.
DATA REQUIRED FOR FURTHER PROCESSING. IN THE MASS SPECTROMETER
THE PEAK REFLECTS RELATIVE ABUNDANCE OF THE PARTICLE AND ITS 
LOCATION IN THE SWEEP RFLECTS ITS MASS NUMBER. IN GENERAL THIS 
IS TRUE FOR OTHER TYPES OF SWEEPING SPECTROMETERS. IN 
THE COULTER PARTICLE COUNTER,PEAK AMPLITUDE REPRESENTS THE
DIAMETER OF A SINGLE PARTICLE  AND USUALLY IS LOST INFORMATION
THE TECHNICON AUTOANALYZER OUTPUT IS A PEAK REFLECTING 
CONCENTRATION OF TEST MATERIAL. IN THE ANIMAL LABORATORY PEAKS AND VALLEYS
MAY REPRESENT COMPONENTS OF BLOOD PRESSURE,THE ELECTROCARDIOGRAM,
PHASES OF FLOW.   AUTOMATIC LOCATION OF PEAKS IS AS ASSET
TO ANALYSIS OF LABORATORY OUTPUT.

IT HAS BEEN DEMONSTRATED THAT A DIGITAL COMPUTER CAN PROCESS ANALOG
SIGNAL. THE SYSTEM IS USUALLY EXPENSIVE TO ACQUIRE, PROGRAM
AND MAINTAIN.

WE PROPOSE TO MAKE THE BEST USE OF DIGITAL COMPUTERS BY REDUCING
THE REAL TIME LOAD, SIMPLIFYING THE PROGRAMMING TASK, AND FITTING
THE PROJECT TO SMALL SYSTEMS. 


FOLLOWING IS A REWRITE OF THE PAPER BEGUN 7/3/73
	WHEN LOW COST MULTIPLE DIAGNOSTIC TESTS WERE MADE AVAILABLE
VIA AUTOMATED TRANSPORT TO ANALYTIC INSTRUMENTS WE WERE FACED  WITH
THE NEED TO PROCESS THE OUTPUT CHARTS AND PLACE THE RESULTS IN 
 APPROPRIATE REPORTS. THERE ARE A VERY LARGE NUMBER OF
AUTOMATED DIAGNOSTIC TEST SYSTEMS IN THE FIELD AND 
THEY REPRESENT THE MAJOR APPLICATION  OF THE PEAK DETECTOR
WHICH I AM PRESENTING TO YOU THIS MORNING.  THE PEAK DETECTOR
PREPROCESSES THE REDUNDANT  ANALOG OUTPUT OF AUTOANALYZER
TYPE INSTRUMENTS AND DELIVERS ONLY A SINGLE VALUE PER PEAK
AS OUTPUT FOR REPORT GENERATION. THE DIGITAL COMPUTER  IS THE 
PREFERRED MEANS FR GENRATIONG A REPORT AND COLLATING PATIENT
FILES BUT IT IS A VERY INEFFICIENT USE OF THE COMPUTER TO
LOAD ITS LIMITED MEMORY WITH THE PROGRAMS FOR PICKING PEAKS FROM
HUNDREDS OF DATA POINTS.