Scans like this one can be difficult for doctors to interpret

The program mimics the way that leading surgeons diagnose tumours, interpreting patterns of information using magnetic resonance spectroscopy (MRS).

MRS scans analyse the pattern of chemicals in tumours and calculate whether they are cancerous.

Professor John Griffiths - "you won't even have to touch the patient"

Doctors have found it difficult to interpret the scans accurately, however, so most of them resort to surgery to remove a small sample of the suspected tumour and examine it by microscope.

Not only is this surgery costly, but one in 100 patients die as a result of the operation and 3% suffer serious complications.

Each scan produces a 'signature'

The new program learns the 'signatures' of various types of tumours and compares them with a patient's scan.

A European trial indicates it is more accurate than most doctors and almost good as leading specialists.

Professor John Griffiths of the Cancer Research Campaign, who helped to develop the program, said: "We believe we can improve on the accuracy rate of 80% to make it even better."

Professor John Griffiths from St George's Hospital, London

He added that the results showed that MRS scans could eventually prove better at assessing and diagnosing tumours than operating on people to remove tissue, and he is hopeful that the program could be adapted to diagnose other forms of cancer.

The robot doctor is one of several new MRS scan programmes under investigation.

Another, developed by Dr Rosemary Tate of Birkbeck College in London, uses maths to interpret MRS scan patterns and is also said to be 80% accurate.

Push of a button

Both programmes are mentioned in the Cancer Research Campaign's Scientific Yearbook.