Computers in Medicine

From a dissertation read before the Society on Friday, 4th November, 1966 The subject of this dissertation is Computers in Medicine and even those who have had nothing to do with these machines will be unable to ignore them in the very near future. A brief account of how they work is given here, followed by the description of a few of their applications in Medicine. In fact, learning to programme the machine is very simple, and the University Computer unit runs a special course three times a year for this purpose. Many people think of the computer as something between a glorified adding machine and a sort of god that can do anything, whereas in fact the truth lies somewhere in between.


Richard D. Turner
From a dissertation read before the Society on Friday, 4th Novem ber, 1966 T h e subject of this dissertation is C om puters in M ed icine and even those who have had nothing to do w ith these m achines w ill be unable to ignore them in the very near future.A brief account of how they work is given here, followed by the description of a few o f their applications in M ed icine.In fact, learning to program m e the m achine is very sim ple, and the University C o m p u ter unit runs a special course three tim es a year for this purpose.M any people think of the com puter as som ething between a glorified adding m achine and a sort of god that can do anything, whereas in fact the truth lies som ewhere in between.

PROGRAMMING
It should be emphasised that usually the programm er need know very little m athem atics.T h e program m ing language is a standard one, and the person who has a particular skill, say in m athem atics, can write out a program m e for carrying out any particular procedure.H e m ight, for instance, devise an ingenious pro gram m e for evaluating the square roots of com plex num bers.H e could publish this program m e in a magazine, and then anyone else w ishing to do the sam e thing as part of another program m e could copy this out word for word w ithout having any understanding of the m ethod involved.T h ese " little program m es" are term ed A lgorithim s, and the index o f these is by now very extensive.
T h ere are several features of any com puter which make it an extrem ely versatile m achine, b u t three should b e stressed in particular -(1) it can carry out sim ple operations exceed ingly fast; fo r instance, the m achine can add two digits together in about a m illionth of a second, and all m ultiplic ation is done by repeated addition.(2) it can follow a drain o f instructions.T h u s to work out the equation: x = (a x b) + (c -d) it is necessary to do the follow ing: m ultiply a by b subtract d from c add the first result to the second print the result and call this x and the com puter can follow these instruc tions in that order.
(3) it can " choose" between follow ing one set of instructions or another according to the result of som e interm ediate calculation.T h u s if one were trying to find w hether a num ber were divisible by 3, 5, or 9 one could tell the com puter:-if the result = a w hole num ber then print the result, otherwise tr y th e next one.If there is no " next one" then print o.
These properties look simple, but enable a very wide range of problems to be tackled.The programme often looks long winded on paper, but the machine is extremely fast.
The following diagram shows the essential features of any digital computer.

( on paper tape )
The instructions and data are usually coded in the form of punched paper tape.Each key depressed on a special typewriter produces a different combination of holes across the tape, and in the machine these holes can be " read" by means of a light shining through them onto a row of photo-electric cells (the input device).Most machines will accept magnetic tape and punched cards as well.
All instructions for the machine are written in a special programming language (usually Atlas Autocode in Edinburgh) and this is very like English; but when the language compiler receives a word such as "add" a + b it will in turn switch on the various devices inside the computer required to add these two digits together.
The store is the memory of the machine, performing a function similar to that of a taperecorder, and can be considered as a large number of boxes holding numbers, which can be transferred to and from the calculating unit or printed out as directed by the instructions.The output device is usually a " Line Printer" although paper tape is sometimes used and this can be fed back through the special typewriter to produce a printed result.

USE OF the computer
To illustrate the use of the machine a pro gramme can be considered.For instance, it might be required to find the number of times a letter occurs in a sentence.Regarding the store as a series of boxes, a Flow Diagram for this programme would appear as follows:- (1) Begin (2) put the letter to be counted into box 1 (3) start reading the letters of the sentence, one at a time, into box 2 (4) if the contents of box 1 = box 2 then add 1 to box 3, otherwise go on to the next letter (5) if the last letter of the sentence has been reached print out the total in box 3 (6) stop.
These instructions can easily be written out in Atlas Autocode, followed by the data which in this case would be the letter to be counted, and the letters in the sentence.Now this programme needs very little alter ation to count the number of times a particular word, or particular phrase, occurs in an entire book, and it was this method which was used to determine whether in fact Paul wrote all the chapters attributed to him in the Bible since any writer characteristically uses particular phrases with a particular frequency.Another possible use of this programme might be to search through vast amounts of literature for a particular reference such as, say, "Acute Tubular Necrosis" .This idea could be extended to make the machine print out a list of refer ences of articles which contained both the phrase "Acute Tubular Necrosis" and the phrase " Hypertension" thus providing an inter esting method for extracting any desired information from any amount of literature Others have tried to apply similar programmes to the automatic translation of language, but so far without very much success.
Another idea which has been put forward is the Automatic Library Facility (ALF).Suppose Dr. A. has certain items of information 11, 12, i:i and 14 (which may be symptoms, signs, or results of tests, etc.) concerning a patient, but that he requires more information on the differential diagnosis than his own knowledge or experience affords.Normally he would consult his own colleagues and the available literature but it is now proposed that in addition he consult A LF. Asked for the very first time by Dr. A. what diseases are associated with 11, 12, 13, and 14 , A L F will answer none, which of course is no great help to Dr. A. who can him self think of two such diseases D 1 and D 2. A LF stores these two.When next asked a similar question by Dr. B., A L F obviously produces D 1 and D 2, one of which is news to Dr. B., but who notes that A L F has omitted a third possibility, D3.In effect, A L F now "knows" the associations of Drs. A. and B. and clearly after 100 doctors have made their enquiries, A L F 's suggestions w ill be o f use to any doctor who does not know as m uch as all th e previous enquiries put together.
Furtherm ore, as know ledge about individual diseases accum ulates, A L F could also indicate an exhaustive set of other differentiating signs, tests, etc., and if told of the m ethods o f treat m ent used together w ith the results obtained, the com puter w ill be able to com pare these and offer advice to others on the best treat m ent available to date.
Y e t another im portant use o f the com puter organisation described is in drug testing.If it was m andatory to report the prescription of any new drug, together w ith any sym ptom s appearing which were not present before the drug was prescribed, then it would be surprising if side-eff ects were not recognised very soon.In this context it is w orth rem em bering that even after thalidom ide, the problem o f con genital abnorm ality rem ains, and a system atic, statistical analysis o f all relevant data could be a pow erful research tool.
T h ere is a continual search for tests which are absolutely diagnostic of specific diseases, and the com puter can b e o f value in this field also.F o r instance, although there m ay b e no single test available, it is often the case that a certain set of sym ptom s, signs, etc., m ay b e absolutely diagnostic and statistical analysis o f all the factors relevant to each disease m ay indicate cases where this is so.T h e principle can be extended to include assessment o f the signific ance o f th e various waves produced by E C G or E E G m achines.T h e n in certain acute disease, such as m yocardial infarction, meaurem cnts of B .P ., pulse rate, or blood chem istry m ay be carried out by autom atic equipm ent already available, and the resultant data analysed by the m achine which would indicate the treatm ent required, or m igh t even initiate such treatm ent in som e cases.T h e advantage o f the m achine when prescribing a routine of treatm ent is that the needs o f th e individual p atient are catered for, w hich should be better than using a standard routine.
T h e re are m any other uses for the com puter w ithin the hospital and an exam ple is as an aid to the bacteriologist.
B acteriological identification proceeds by subm itting a specim en to a num ber o f tests, deciding on th e basis o f th e results w hich are the m ost likely possibilities, then carrying out further tests to differentiate between these, and so on until identification is established to the bacteriologist's satisfaction.A com puter pro cedure has been developed to assist the process as follow s: the results 'phoned in to the com puter centre m ay be as in F ig .1.

C o m p uter
In p u t lactose = acid sucrose = acid indole = negative m otility = negative H :S = negative U rea = negative A t the centre these are typed out, and the resulting piece of punched tape entered into the com puter.W ith in a few m inutes it w ill print out the m ost lik ely bacteria in order of preference, and also the next set o f tests which will m ost efficiently discrim inate between these.(See F ig 2).
F .86 •79 relative likelihood relative discrim inating power In actual cases it has been found that the com puter has agreed w ith the bacteriologist in all im portant respects, and has generally established the diagnosis using few er tests.O f even greater benefit to the patient, however, is th e fact that the com pu ter keeps up to date w ith all the latest developm ents in the field.

C OMPUTER CENTRES
It should be noted that in m ost cases, the m achine spends far less tim e calculating than the operator takes in deciding w hat to do w ith the results.A T im e Sharing technique is usually used so that data is fed in to sm all m achines, perhaps one in each hospital event ually one day, and a large central com puter turns its attention to these to suit itself, thus saving tim e and m oney.T h e central m achines are being set up at various places in the country for general U niversity use, by the governm ent, and m ethods are being developed by w hich these m achines can com m unicate with one another also, which w ill have vast potential one day.
A s far as M ed icin e is concerned the sam e device w hich can com m unicate with w orld diagnostic or inform ation networks could be used to calculate nursing schedules or teach-ing tim etable, or m ay com pare the vital statistics o f one hospital w ith another to indicate how an epidem ic is spreading.
T h e next difficulty is that as the requirem ents get m ore com plicated, so program m ing the m achine takes longer, and several centres are trying to devise systems in w hich the m achine programmes itself.As an example, various attem pts have been made to write a program m e for the com puter to play chess.For a m an to tell the m achine w hat to do in every possible situation in advance w ould take a very long tim e. H ow ever the m achine can survey the positions of all the m en on the board and then try a m ove at random.If this results in im m ediate gain, or if it eventually wins the gam e, it w ill try the same moves for each tim e the particular com bination o f men recurs in future games.If it loses, it w ill rem em ber w hat its opponent did, and will use his moves in future instead.Analysis of gam e theory in this way has led to im portant advances in other sciences, and may well be of use in the m edical field in the future.
A ltogether the com puter m ay w ell provide the spectacular changes in M ed icin e in our generation that antibiotics and bacteriology provided for our forefathers, and tim e spent acquiring som e know ledge of how it works w hile at U niversity is m ost unlikely to b e tim e wasted.1. History of bruising, em otional lability, increasing w eight and lethargy.2. Plethora.

D iastolic hypertension.
W h a t is the diagnosis?H ow would you confirm it?
Set by Robin B. L. Ewart, M.B., Ch.B., B.Sc. (answer on page 44) S u b je c t : C .S ., married, fem ale, aged 33.Housewife.P ast H istory : 1. Usual childhood illnesses.2. Pneum onia on three occasions between the ages of 3 and 5 years.3. R heum atic fever, aged 9 years.4. Ton sillectom y, aged 10 years. 5. A ppendicectom y, aged 16 years.6. T h e patient had tw o normal pregnan cies, aged 23 and 24 years.7. Perforation of duodenal ulcer, aged 25 years.-Gastroenterostom y. 8. G radually progressive exertional dysp noea began, aged 27. 9. T w o miscarriages, aged 28, follow ed by tubal ligation on the grounds of rheum atic heart disease.10. M itral valvulotom y successfully carried out, aged 29, w ith relief of symptoms.11.Recurrence o f classical acute rheum atism, aged 30.F ollow ing discharge from hospital, th e pat ient seemed initially to b e well b u t 2 1 /2 years later was readmitted w ith the follow ing com plaints -31 1. T en d en cy to bruise on m inor trauma, -1 year.2. M arked em otional lability -1 year.3. Progressive increase in w eight am ounting to 9 lbs. in the previous six m onths.4. Increasing lethargy -3 m onths.Social and Fam ily H istory : N o t relevant.Exam ination : Plum p, plethoric w om en, looking older than her 33 years.M arked bruising o f all four lim bs was evident.C .V .S .Pulse 102.Regular in tim e and force.B .P. 180/110.Auscultation of the heart revealed the classical signs of mitral sten osis and incom petence.T here was no sign of cardiac failure.All other systems essentially negative to full exam ination.