Robot that performs work of 15 cancer doctors

A ‘ROBOT’ doctor called Watson has been pitted against a panel of 15 leading medics – and found to be just as good as them at treating cancer.

In a development that will send a chill down the spine of health professionals everywhere, the IBM-designed computer program worked out the best way to treat patients in a fraction of the time that it took the highly paid specialists.

Working together, 15 consultant oncologists took 12 minutes on average to recommend a treatment plan for each patient. But it took the supercomputer, dubbed Watson For Oncology, just 40 seconds.

Man and machine were each given notes from 638 former breast cancer patients to assess – minus one vital piece of information: the treatment plan their doctors had chosen.

In nine out of ten cases of early-stage breast cancer, both human and artificial intelligence came up with identical recommendations, according to the results of the study.

Watson, which works in a similar way to mobile phone personal assistants such as Siri, has been ‘taught’ to weigh up evidence by doctors at the world-renowned Memorial Sloan Kettering Cancer Centre in New York.

It has already assimilated 200 medical textbooks, the case files of four million patients, and has access to cutting-edge research.

Cancer expert Dr Andrew Nordon, deputy chief health officer at IBM Watson Health, said: ‘It doesn’t get tired, it isn’t susceptible to human biases, and it has a limitless capacity.’

But Watson and the human panel had to agree to differ when it came to treating more advanced disease, where breast cancer had spread. In such cases they disagreed 55 per cent of the time. Dr Nordon also conceded: ‘It can’t look a patient in the eye and understand their emotional state – yet.’

Nevertheless, Watson is already being used in a number of leading US cancer centres, and some in India and Thailand.

But top UK cancer doctor Professor Jack Cuzick remained unfazed. He said: ‘Real doctors will always be one step ahead of a machine that can only review data retrospectively.’