ict in health

7/31/2019 ICT in Health

1/18

AS ICT1 Information Systems ICT in Health

COMPUTERS IN HOSPITALS

Mr. John Martin arrives at the hospital to see his patients. He will examine therecords of their vital signs of life and decide if anyone is well enough to leavethe Intensive Care Wards and to return to the normal wards. He will then decideupon his operating list for the day, as all neurosurgery operative patients need

intensive care post operatively. This is in the day to day running of a busyhospital department.

The hospital being in close proximity to the M4 deals with numerous cases daily.It now boasts a helipad that allows chronically ill patients to be transferredquickly to the A&E department.

Mr. Martin will use all the advantages that ICT can give him in the care of hispatients. He will use ICT in numerous different forms as he gather theinformation needed to help his patient and to ensure that he cares for them in

to the best of his ability.

ICT is at the heart of the patient care. It is developing in all exciting and diversemethods in the quest to help care for ill people. Let us look at some of the waysthat this occurs.

There are two main areas of computer use in hospitals:

Basic ledger systemUse for patient care

Ledger System

This deals with the day-to-day purchases that the hospital needs to keep goingon a day-to-day basis. It covers everything that is bought by the hospital andaccounts for their purchase and their use by a particular department. Alldepartments have budgets and these are then managed by the computersystem.

The finance system also manages the payroll needs of the hospital. This is the

basis by which health authority pays the salary of each individual.

Advantages

Easy to trace the spending of individual areas Cheaper sourcing of materials and items when buying bulk Budgets are set early on so that they can be adhered to Allows overall monetary control to be decentralised

Disadvantages

Expensive patient care has to be balanced with the budgetary controls Epidemics can not be planned for New procedures tend to be expensive

1


2/18


Drug bills are expensive

For Patient Records

The creation and management of lists.

Patient personal records are not kept on computer. There are many issues as towhy this is and they will be dealt with later on. At present, patient records aretranscribed manually and these records are then held in the Medical RecordsDepartment. A computer is used to manage the list of appointments for clinicsthat go on daily at the busy hospital. Waiting lists are produced for each clinicand the names of the patients on the list are then processed so that their notesare delivered to the correct clinic ready for the days work. The notes caninclude a record of previous visits, previous consultations and diagnoses andinclude also results from tests.

So large is this delivery that a small truck runs around the hospital continuallyall day long being loaded with notes for the clinics and for the return journey toMedical Records.

A patient is assigned a unique number when they enter a hospital, the keyfieldin a database This unique number is only used in that hospital and is nottransferable to any other, even those within the same health authority or trust.The computer held patient records does not carry any detail of their complaintor diagnoses but holds only their doctors name, their GPs name, the date oftheir appointments and the consultants name that they are under. (The name

of the doctor that is treating them may differ from their consultants name)

The computer system allows for case note tracking and as an administration toolfor the overview of lists and details. When a list appears, the medical recordsteam prepare the appropriate notes and test results for the individual and theseare then delivered to the designated ward for admission or to the clinic that theyare attending that day

The use of the keyfield allows direct access to the patient record. The patientsrecords are kept up to date manually and the Medical Records teams type these

up. Detailed notes, such as diagrams by the doctor/surgeon are left in recordsand all printed work is transcribed.

The different areas of the hospital have different uses of IT. The systemsdeveloped are unique in their specialism so that they perform the taskdesignated for them.

The systems continue to evolve, becoming more sophisticated as they aredeveloped and enhanced. They become more sophisticated in what they do,

clever in detecting the tasks prescribed to them. This clearly demonstrates thecontinual development of ICT.

The Haematology Department (Blood)

2


3/18


One computer system is used to administer blood tests and to track blood thatis transfused to patients.

Transfusions.

An ill patient is brought into hospital needing an operation. The patients detailsare recorded and using this information, a bracelet is produced with the keyfieldnumber printed on by barcode. The use of the barcode is to make the input ofinformation as correct as possible. The scanner used reduces the humaninvolvement and the chance of error there in. The barcode is used to track thetransfusions that may be given to the person.

On the unit of blood, there is another unique barcode, which can trace the bloodfrom who gave it. This allows the tracking of the blood as it comes into thesystem and can trace the individual that gave the blood. This system relates to

problems that have arisen over the last few years, especially the CJD (Mad Cowdisease) and how that could be spread by transfusion.

When a unit of blood is despatched to a patient, the blood transfusiondepartment scans it and the appropriate barcode is printed out at thedepartment and stuck on to the unit of blood. When it reaches the patient, thescanned barcode on the patients bracelet and the scanned barcode on the unitof blood have to match before the blood can be administered as a part of atreatment or operation.

Part of the bracelet printed At Morriston Hospital

This system used is called ISBT 128 and is similar to the ISBN system for thecataloguing of books. The system has the following advantages

Advantages

Blood can be tracked from donor to patient Ensuring that the correct blood is given to the correct patient, cutting

down the chance of mistake. Cutting down the chance of cross contamination

So blood is tacked from the donor to the patient. It can be detected in thesystem and its bar coding reduces the human error that may occur in thesystem.

Other systems in Haematology are aimed towards the computerisation of thetests that take place on the blood of patients. In the Pathology laboratory, theuse of ICT enables the diagnosis of blood disorders and biochemical problems.

3


4/18


This may be the counting of red blood corpuscles, the analysis of thecomposition of the blood etc. and all these are carried out automatically. Themachines are regularly calibrated and checked to ensure their accuracy byoutside bodies. (I think thats a PUN)

Results from these tests are transferred to the ward and the doctor that ordered

them via a computer system. This system uses a LAN to communicate theinformation to the doctor and the correct ward as soon as it is available. Thisspeeds up the return of the appropriate results to the doctor to aid the diagnosisof the patients problems as quickly as is possible.

The automation of the tests reduces the chance of human error by using amachine to analyse the samples. It reduces the need for highly trained staff asfar as possible. There is an issue here with the over-dependence with aparticular system. If this system goes down, what would happen to the test thatare needed immediately

The only place where this is not possible is in cytology tests where the samplehas to be viewed via a microscope. The trained laboratory technician is lookingof a change in the cell structure or the presence of wrong type of cell. Thesemay be rogue cells or pointers towards future problems. This has to be done bya trained pathologist and it only a trained person who can be entrusted with thiswork. No computer program is yet in place to do away with the trainedoperative in this field.

Hospital Communication The use of WAN

To communicate information between hospitals for example from BronglaisHospital in Aberystwyth to Morriston Hospital in Swansea closed network isused. This allows x-rays and scans to be sent by a doctor to a specialist forthem to share their opinion. This gives the doctor a second opinion by aspecialist in that field. The closed nature of the network ensures theconfidentiality of the sent material. All hospitals in Wales are linked together andthere is also a gateway to English hospital. It is a system that ensures thecorrect and best advice for that patient. The message can be sent quickly to itsdestination and the information can be discussed whilst looking at the image.

4


5/18


Intensive Treatment Unit (ITU)

The area of Intensive Care and Treatment in one with the use of ICT at its core.Patients can be admitted after trauma (e.g. an accident such as car crash orindustrial incident) or post operatively where they may have undergone majorsurgery. Some patients are also referred when an illness or condition escalates

in seriousness.

All patients in ITU are in a one to one situation with a nurse to every patient.Sensors linked to computers and alarms constantly monitor the patient. Thereadings of the sensors is recorded by computer but legally at the moment, thenurse records the maximum and minimum readings manually. These records arekept for all patients and are shown top the doctor in charge of the case.

Different sensors can measure a number of different body functions such as

Temperature Pulse Blood pressure Central venous pressure Heart rate Blood gases e.g. oxygen in the blood Gases as a breakdown of breath Brain monitoring measuring brainwave activity Continual ECG heart monitoring Fluid level testing Inter cranial pressure (pressure in the skull)

With dedicated computers and sensors for these aspects mentioned above,there is the continual monitoring of the patients, 24/7. Alarms are set forextreme readings allowing nurses and doctors react to each condition asnecessary.

The data can then be turned into information as it is represented as graphs orwhen the maximum or minimum readings are read. It allows doctors to spottrends in patient care and to look at the effect of administration of drugs and

dosage. This is only truly possible with the computer records where certaintimes can be examined in greater detail enabling closer examination at definitetimes in the patients treatment.

Doctors can then act on the input that they are receiving from the equipment.For example, the central venous pressure gives the anaesthetist the informationto prescribe more or less fluids to ensure the maintenance of fluid levels.

All this gives the best possible care for the ill patient. It gives the patients thebest possible chance for recovery. An area though that cannot be quantified by

sensor or probe is the measurement of pain and that is a value judgement thatcannot be placed on computer record.Quantifying Measurements for Input to a Computer

5


6/18


Apache

All patients entering ITU are given an APACHE score. This is a raw diagnosis oftheir illness quantified so that it can be measured. The figure is used to see howa patient improves and gets well. The details of all patients in ITUs are sent onto ICNARC (Intensive Care national Audit Research Centre). Here the centre

looks for areas of good practice and specialism. If a particular centre isproducing good results then the good practice of the centre may be shared withothers.

There is has to be the quantifying of the measurements of illness so that it canbe put onto a computer database and compared to other centres across thecountry. Whilst various factors can be easily measured such as temperature andheart rate, others become difficult to quantify such as the measurement of pain.

Diagnostic Tools

The MRI Scanner

The Magnetic Resonance Imaging Scanner is a radiology technique that usesmagnetism, radio waves and a computer to produce images of the bodystructure. A huge circular magnet is in the tube that surrounds the patient.When the magnet is switched on, the detection of change in magnet resonanceis picked up by radio waves and this change can be modelled by a computer toproduce an image. Different tissues change the resonance of the magneticwaves. The image and resolution of the image produced is quite detailed and

can detect tiny changes in the body. By changing the reading of the magneticresonance and by measuring the change, it can focus the image on differentaspects of the patient, producing some quite spectacular results.

This enables the doctors to investigate problems that could only be looked atpreviously by surgery. For example, the surgeons can now examine the inside ofthe heart and how it is functioning. The inner part of the brain can be examined.

By changing the resonance of the scan, different tissues and other areas such asblood vessels, tissue, bone or ligament can be seen clearly in the resultingpicture. Then neurosurgeons can investigate the integrity of the spinal cord after

6

The scan is made up of a numberof slices through any section ofthe patient. The distance these

slices are apart can be set by theoperative to the request of thedoctor. It can show up quiteclearly any part of the body.


7/18


trauma; brain aneurysms and tumours can be seen clearly. Contrast agents canalso be introduced to increase the accuracy of the images.

Accurate pictures can be taken of joints that show up the soft tissue as well asthe bones of the body. The heart and aorta can be clearly seen as well as glandsand organs, all these without surgery.

Sir Peter Mansfield and Paul C. Lauterbur were the two individuals whodeveloped MRI scanning. The American Lauterbur discovered the possibility ofproducing a two-dimensional picture by variations in the magnetic field.Mansfield, from Nottingham, showed how the emissions of the magnetism couldbe mathematically analysed. He also demonstrated how fast imaging could beachieved though it took more than a decade for his projections to becomereality. For their work in developing this scanner, both were awarded the NobelPrize for medicine in 2003.

Parallel processing and distributed computing

In distributed computing, a number of computers would process its own share ofthe work independently. This is when the computers that are linked together areall set onto the same task at the same time. An example was in Canada wherenumerous computers were set a task overnight via the Internet by a university.Their combined computing power allowed them to tackle a problem too large foran individual computer to take on.

Parallel processing is the use of many processors in a single computer. This

allows the processors communicate with each other much more quickly. Multipleprocessors make the writing of programmes much more difficult and this is whythey are not widely used.

Some MRI scanners use two 930-MHz processors linked together and extremelylarge hard drives (6 terabytes and larger) to capture the resultant data. Theprocessing power will be used to put the scans together into images, which canbe understood by the doctor.

The memory of these machines is backed up every other day, the files being

written to optical discs, these then being stored. It allows for a large amount ofdata to be backed up and catalogued. When a patient returns for check-up, anew scan can be done and the two images, one from before the treatment andone after treatment can be compared. This can also be examined when thepatients return after a few years on a matter not connected or otherwise. Itgives the doctor a much broader picture of the patients health and help with anaccurate diagnosis and treatment.

All scans are backup on optical disk every three days. The use of an optical diskis to future protect the images.

The Future

7


8/18


The role of Director General of the NHS Programme for ICT is a vital one. Thereis a need to take healthcare into the digital generation.

There are many avenues that are being investigated and acted upon. One isElectronic Patient Records (EPR). This is the venture to put all patients recordsonline so that all hospitals have access to them, If you were taken into any

hospital anywhere in Britain they could call up your notes and see if you are forexample on any medication at the moment. This can only lead to betterdiagnosis of patients problems and improve the chances of survival,

This would allow the sharing of information between hospitals, between trustsand health authority, something that does not go on now. It would greatlyimprove the care a patient receives at an A & E department.

Problems associated with the introduction of such a system would be thecompatibility to all. Large projects such as this are notorious for running over-

budget and time and the need for the system to be robust, user friendly andfuture proofed create an immense task. Security of Information is also a majorconcern as well as the updating of all patient records. Many issues have to bedealt with.

These plans would also make it possible to book appointments with the doctorover the Internet and would be able to look at their own notes also. Patientswould also be able to choose where they went for an appointment and to selectthe practise they attend for a consultation. Your personal notes are locked in afiling cabinet somewhere and at present there are something like 660 million

pieces of paper in the NHS system and a vast majority have been typed 2 or 3times. Streamlining this would cut down on a lot of waste and repetitive work.

The computerising of the patient records would also create a huge database ofall the people in Britain. This could also allow an epidemiological investigation totake place. Peoples records could be compared to see what treatment workedand what other factors there may have been for some complex illness. Trendscould be spotted at their early stage and remedial action taken to stop themquickly. This information would be available no matter where the hospital was orwhat the condition of it is.

A database such as this can also be seen as a step towards a national identityprogramme where to get treatment one would have to proof of nationality andproof of residence in this country. You would have to exist on the databasebefore one can be dealt with. Immigrants, legal or otherwise would have toprove their adherence to these rules.

There are many problems and a lot of work has to go into their developmentbefore they can be achieved. Numerous issues need clarification and manyavenues have to be explored before completion. There is though a great

improvement that can be made and a huge streamlining of the bureaucracy ofthe NHS letting managers target the money spent in other directions.

EXPERTS

8


9/18


Artificial Intelligence

Software is now being developed which will enable computers to learn andreason. e.g. Some chess game programs get better the more they are played -the computer remembers 'when I made that move, I lost...therefore find an

alternative'.

AI is difficult to define. Alan Turing (a prominent mathematician) developed asimple test (1950) to determine if a computer possessed intelligence:

Suppose there are two identical terminals in a room, oneconnected to a computer, and the other operated by aperson. If someone using the two terminals is unable to tellwhich is connected to the computer and which is operated bythe person, then the computer can be credited with

intelligence.

AI research includes

language processing - understanding and speaking languages aswell as humans. computer vision - recognising and analysing objects.

Neural networks.

The current technology for this is only in its infancy. Nothing to do withcomputer networks, neural networks try to mimic the way that the humanbrain works (neurons, synapses etc).

Work on neural networks is being carried out in the field of image analysis,pattern analysis, financial trends etc.

Aspirin is a language used in neural networks (Using a MIGRAINESinterface!).

Requirements

HARDWAREParallel processors with powerful capabilities to handle numbers very quickly

SOFTWARELanguages such as OCCAM

How does a parallel processor handle an instruction?

Instruction is split up and each processor does part of it

e.g. 4 processors adding up 8 numbers

9


10/18


P1 P2 P3 P41+2 3+4 5+6 7+8P1+ P2 P3 +P4

Result of above

Expert Systems

An expert system is a computer system that emulates the decision-makingability of a human expert.

A knowledge-based system that attempts to replace a human 'expert' in aparticular field.

It diagnoses problems and gives advice on the cause of those problems. They

can also give advice on solutions.

Components of a rule-based expert system

A typical rule-based expert system integrates

1. A problem-domain-specific knowledge base that stores the encodedknowledge to support one problem domain such as diagnosing why a carwon't start. In a rule-based expert system, the knowledge base includes

the if-then rules and additional specifications that control the course of theinterview.

2. An inference engine a set of rules for making deductions from the dataand that implements the reasoning mechanism and controls the interviewprocess. The inference engine might be generalized so that the samesoftware is able to process many different knowledge bases.

3. The user interface requests information from the user and outputsintermediate and final results. In some expert systems, input is acquiredfrom additional sources such as databases and sensors.

An expert system shell consists of a generalized inference engine and userinterface designed to work with a knowledge base provided in a specified

format. A shell often includes tools that help with the design, development andtesting of the knowledge base. With the shell approach, expert systems

10


11/18


representing many different problem domains may be developed and deliveredwith the same software environment.

There are special high-level languages used to program expert systemse.g. PROLOG

The user interacts with the system through a user interface that may usemenus, natural language or any other style of interaction). Then an inferenceengine is used to reason with both the expert knowledge (extracted from ourfriendly expert) and data specific to the particular problem being solved.

The expert knowledge will typically be in the form of a set of IF-THEN rules. Thecase specific data includes both data provided by the user and partialconclusions (along with certainty measures) based on this data. In a simpleforward chaining rule-based system the case specific data will be the elementsin working memory.

How an expert system works

Car engine diagnosis

1. IF engine_getting_petrolAND engine_turns_overTHEN problem_with_spark_plugs

2. IF NOT engine_turns_overAND NOT lights_come_onTHEN problem_with_battery

3. IF NOT engine_turns_overAND lights_come_onTHEN problem_with_starter

4. IF petrol_in_fuel_tankTHEN engine_getting_petrol

There are three possible problems with the car:

problem_with_spark_plugs, problem_with_battery, problem_with_starter.

The system will ask the user:

Is it true that there's petrol in the fuel tank?

11


12/18


Let's say that the answer is yes. This answer would be recorded, so that theuser doesn't get asked the same question again. Anyway, the system now hasproved that the engine is getting petrol, so now wants to find out if the engineturns over. As the system doesn't yet know whether this is the case, and asthere are no rules that conclude this, the user will be asked:

Is it true that the engine turns over?Lets say this time the answer is no. There are no other rules which can be usedto prove ''problem_with_spark_plugs'' so the system will conclude that this isnot the solution to the problem, and will consider the next hypothesis:problem_with_battery. It is true that the engine does not turn over (the userhas just said that), so all it has to prove is that the lights don't come one. It willask the user

Is it true that the lights come on?

Suppose the answer is no. It has now proved that the problem is with thebattery. Some systems might stop there, but usually there might be more thanone solution, (e.g., more than one fault with the car), or it will be uncertainwhich of various solutions is the right one. So usually all hypotheses areconsidered. It will try to prove ''problem_with_starter'' but given the existingdata (the lights come on) the proof will fail, so the system will conclude that theproblem is with the battery. A complete interaction with our very simple systemmight be:

System: Is it true that there's petrol in the fuel tank?

User: Yes.

System: Is it true that the engine turns over?

User: No.

System: Is it true that the lights come on?

User: No.

System: I conclude that there is a problem with battery.

Note that in general, solving problems using backward chaining involvessearching through all the possible ways of proving the hypothesis,systematically checking each of them.

Questions

1. ''Briefly describe the basic architecture of a typical expert system,mentioning the function of each of the main components.''

2. ''A travel agent asks you to design an expert system to help peoplechoose where to go on holiday. Design a set of decisions to help you give

advice on which holiday to take.''

Expert System Use

12


13/18


Expert systems are used in a variety of areas, and are still the most populardevelopmental approach in the artificial intelligence world. The table belowdepicts the percentage of expert systems being developed in particular areas:

Area Percentage

Production/Operations Mgmt 48%

Finance 17%

Information Systems 12%

Marketing/Transactions 10%

Accounting/Auditing 5%

International Business 3%

Human Resources 2%

Others 2%

Medical screening for cancer and brain tumours

Matching people to jobs Training on oil rigs Diagnosing faults in car engines Legal advisory systems Mineral prospecting

Medical diagnosis

The computer does not take the place of the doctor but can be used to help thedoctor make decisions.

An expert system would have information about diseases and their symptoms,the drugs used in treatments etc.

A patient is asked by a doctor about symptoms and the replies are input to theexpert system. The computer searches its database, uses its rules and makessuggestions about the disease and its treatments. Sometimes probabilities areassigned to diagnoses.

Mycin was one of the earliest expert systems, and its design has stronglyinfluenced the design of commercial expert systems and expert system shells.

Mycin was an expert system developed at Stanford in the 1970s. Its job was todiagnose and recommend treatment for certain blood infections. To do thediagnosis ``properly'' involves growing cultures of the infecting organism.Unfortunately this takes around 48 hours, and if doctors waited until this wascomplete their patient might be dead! So, doctors have to come up with quickguesses about likely problems from the available data, and use these guesses toprovide a ``covering'' treatment where drugs are given which should deal withany possible problem.

Mycin was developed partly in order to explore how human experts make theserough (but important) guesses based on partial information. However, theproblem is also a potentially important one in practical terms - there are lots of

13


14/18


junior or non-specialised doctors who sometimes have to make such a roughdiagnosis, and if there is an expert tool available to help them then this mightallow more effective treatment to be given. In fact, Mycin was never actuallyused in practice. This wasn't because of any weakness in its performance - intests it outperformed members of the Stanford medical school. It was as muchbecause of ethical and legal issues related to the use of computers in medicine -

if it gives the wrong diagnosis, who do you sue?

Anyway Mycin represented its knowledge as a set of IF-THEN rules withcertainty factors. The following is an English version of one of Mycin's rules:

IF the infection is pimary-bacteremia

AND the site of the culture is one of the sterile sites

AND the suspected portal of entry is the gastrointestinal tract

THEN there is suggestive evidence (0.7) that infection is bacteroid

The 0.7 is roughly the certainty that the conclusion will be true given theevidence. If the evidence is uncertain the certainties of the bits of evidence willbe combined with the certainty of the rule to give the certainty of theconclusion.

Mycin was written in Lisp, and its rules are formally represented as Lispexpressions. The action part of the rule could just be a conclusion about the

problem being solved, or it could be an arbitrary lisp expression. This allowedgreat flexibility, but removed some of the modularity and clarity of rule-basedsystems, so using the facility had to be used with care.

Anyway, Mycin is a (primarily) goal-directed system, using the basic backwardchaining reasoning strategy that we described above. However, Mycin usedvarious heuristics to control the search for a solution (or proof of somehypothesis). These were needed both to make the reasoning efficient and toprevent the user being asked too many unnecessary questions.

One strategy is to first ask the user a number of more or less preset questionsthat are always required and which allow the system to rule out totally unlikelydiagnoses. Once these questions have been asked the system can then focus onparticular, more specific possible blood disorders, and go into full backwardchaining mode to try and prove each one. This rules out a lot of unnecessarysearch, and also follows the pattern of human patient-doctor interviews.

The other strategies relate to the way in which rules are invoked. The first oneis simple: given a possible rule to use, Mycin first checks all the premises of therule to see if any are known to be false. If so there's not much point using therule. The other strategies relate more to the certainty factors. Mycin will first

look at rules that have more certain conclusions, and will abandon a search oncethe certainties involved get below 0.2.

14


15/18


There are three main stages to the dialogue with Mycin:

In the first stage, initial data about the case is gathered so the systemcan come up with a very broad diagnosis.

In the second more directed questions are asked to test specifichypotheses. At the end of this section a diagnosis is proposed.

In the third section questions are asked to determine an appropriatetreatment, given the diagnosis and facts about the patient. Thisobviously concludes with a treatment recommendation.

At any stage the user can ask why a question was asked or how a conclusionwas reached, and when treatment is recommended the user can ask foralternative treatments if the first is not viewed as satisfactory.

A new expert system called PUFF was developed using EMYCIN in the newdomain of heart disorders.

A later version called NEOMYCIN had an explicit disease classification torepresent facts about different kinds of diseases and a system called NEOMYCINwas developed for training doctors, which would take them through variousexample cases, checking their conclusions and explaining where they wentwrong.

Advantages

The computer can store far more information than a human. It

can draw on a wide variety of sources such as stored knowledge frombooks case studies to help in diagnosis and advice. The computer does not 'forget' or make mistakes. Data can be kept up-to-date. The expert system is always available 24 hours a day and willnever 'retire'. The system can be used at a distance over a network. So ruralareas or even poorer third world countries have access to experts. Provides accurate predictions with probabilities of all possibleproblems with more accurate advice.

Some people prefer the privacy of talking to a computer.

Limitations / Disadvantages of expert systems

Over reliance upon computers Some experts could lose their jobs or not be given training if

computers are available to do the job. Lacks the 'human touch'! lack of personal contact Dependent upon the correct information being given. If data or rules

wrong the wrong advice could be given.

Expert systems have no "common sense". They have nounderstanding of what they are for or of what the limits of theirapplicability are, or of how their recommendations fit into a largercontext. If MYCIN were told that a patient who has received a gunshot

15


16/18


wound is bleeding to death, the program would attempt to diagnose abacterial cause for the patient's symptoms.

Expert systems can make absurd errors, such as prescribing anobviously incorrect dosage of a drug for a patient whose weight andage are accidentally swapped by the clerk.

The knowledge base of an expert system is small and therefore manageable--afew thousand rules at most. Programmers are able to employ simple methods ofsearching and updating the KB, which would not work if the KB were large.Furthermore, micro-world programming involves extensive use of what arecalled "domain-specific tricks"--dodges and shortcuts that work only because ofthe circumscribed nature of the program's "world". More general simplificationsare also possible.

One example concerns the representation of time. Some expert systems get by

without acknowledging time at all. In their micro-worlds everything happens inan eternal present. If reference to time is unavoidable, the micro-worldprogrammer includes only such aspects of temporal structure as are essential tothe task--for example, that:

ifa is before b andb is before c then a is before c.

This rule enables the expert system to merge suitable pairs of before-statements and so extract their implication (e.g. that the patient's rash occurredbefore the application of penicillin). The system may have no other informationat all concerning the relationship "before"--not even that it orders events in timerather than space.

The problem of how to design a computer program that performs at humanlevels of competence in the full complexity of the real world remains open.

16


17/18


http://www.aaai.org/AITopics/html/expert.html

Delivering expertise without the expert's physical presence

The scenario we just examined used a telephone to provide remote access to anexpert mechanic. Books and manuals provide other examples of packagedexpertise. Methods for delivering advice without the expert's presence thatinclude a stronger goal orientation include checklists, flowcharts and decisiontables:

AUTO DIAGNOSTIC CHECKLIST

SECTION 11. Does the starter operate?

A. Yes (GO TO SECTION 2)B. No (GO TO SECTION 3)

A checklist for diagnosing why a carwon't start might begin like this. Thebranching nature of the problem could

result in a complex questionnaire.

Graphical representations of diagnosticprocedures like this flowchart provide analternative to complex checklists.

Rule 1 2 3 4

Decision tables can provide proceduralguidance for complex problems. Attributesof the problem are listed in the conditionstub (green) and recommendations orintermediate results in the action stub(yellow). Rules (read vertically) specifythe action to take for any combination ofconditions.

Starter runs? Y Y N

Smell gas? Y N .

Dead battery . . X

Out of gas . X .

Flooded X . .

17


18/18


Representing knowledge in rule-based systems

RULE 1:

Ifthe result of switching on the

headlights is that nothing happens,or the result of trying the starter isthat nothing happens.

Thenthe recommended action isrecharge or replace the battery

RULE 2:

Ifthe result of trying the starter is the

car cranks normally and a gas smell isnot present when trying the starterThen the gas tank is empty with 90%confidence

RULE 3:

Ifthe gas tank is empty

Then the recommended action isrefuel the car

RULE 4:

Ifthe result of trying the starter is the

car cranks normally and a gas smell ispresent when trying the starterThen the recommended action is wait 10minutes, then restart flooded car

Each rule consists of an IF part called the premise or antecedentand a THENpart called the consequentor conclusion. When the IF part is true, the rule issaid to fire and the THEN part is asserted- it is considered to be a fact.

Rule results are often combined to reach a conclusion. The goal of the autodiagnosis is to find a recommended action: what to do to get the car started.Rule 3 tells what to do if the gas tank is empty and rule 2 could prove that thegas tank is empty. If rule 2 fires, rule 3 will also fire and provide arecommended course of action.

The consequent in rule 2 is asserted with 90% confidence. This means that ifthe rule's premise is true, we are only 90% certain that the car is out of gas.Our computer-based expert might be willing to accept this level of confidence tofire rule 3 and recommend an action.

18

ict in health

Documents