The neural computer system, developed to help inexperienced doctors in emergencies, raises the fear of machines making life-and- death decisions over patients.
The computer, designed by Dr Robert Harrison and Steve Marshall, of Sheffield University, and Dr Lee Kennedy, of Edinburgh University department of medicine, is one of the first clinical systems of its kind in the world. A similar machine is being developed in San Diego.
The developers say that it is a 'decision aid' and can provide junior doctors with the experience of specialists. In first tests, it was accurate 90 per cent of the time, compared with 50 to 60 per cent accuracy among doctors. In a small percentage of cases very hard to diagnose, the machine even did better than cardiac specialists.
'This is not really about the majority of heart attacks which are relatively easy to diagnose - the overweight man in his forties with chest pain, who smokes and drinks a lot - but about the 10 or 20 per cent which are very hard to diagnose,' said Dr Kennedy, a consultant physician and senior lecturer.
'It is not just a question of whether or not the machine does better than a doctor. What we have to demonstrate now is that the machine will enable junior doctors to perform better in emergencies.
'These machines do raise anxieties among patients and in the profession as well about doctors losing control of medicine. My belief is that the decision must always remain with the doctors, who do anyway err on the side of caution. The doctor's decision would prevail.'
Dr Harrison said: 'It does raise major ethical and moral questions. It is very important that developments such as this do not in time have the effect of de- skilling doctors. This is why we think that tools such as this should be put in the hands of doctors themselves.'
Dr Andrew Vallance-Owen, head of central services at the British Medical Association, said that it was crucial that doctors maintained their skills and continued to use their clinical judgement.
'Doctors should make notes first of their own diagnosis before they turn to a diagnostic aid such as this,' he said. 'It is very important that doctors continue to be taught the art of putting the information together and coming to a diagnosis based on talking to the patient and their examination of the patient. The machines might not always be available.'
A trial of the computer with heart-attack patients has received ethical approval and will begin shortly in Edinburgh. The doctor who sees the patient will also give the computer information and receive its decision. Later it is planned to mount clinical trials in up to six accident and emergency departments around the country.
Decisions about heart attacks need to be taken quickly, Dr Kennedy said. But results of the enzyme test, which is the best proof that heart tissue has been damaged and that a true heart attack has occurred, take hours while bloodclot-busting drugs need to be given very rapidly.
These drugs are expensive, and could be fatal if given to a patient with a bleeding stomach ulcer, who could have arrived at hospital complaining of sudden pain in the chest. Diagnosis in the elderly can be expecially difficult because of pre-existing conditions and drugs that may disguise true symptoms.
The intelligent computer, called CDS-MI, was developed using information from 500 suspected heart-attack cases, and their suggested and confirmed diagnoses, at the Northern General Hospital, Sheffield.
Dr Harrison, a lecturer in the department of automatic control and systems engineering, said that in some ways the machine was modelled on the human brain's ability to learn from experience, from repetition, and detection and eradication of errors.
'If you tell the system enough times it can adjust until it gets it right,' he said.