The idea that medical treatment in the future will be delivered entirely by automatons may still be in the realm of science fiction, but there’s one area where computers could play an important role in helping out busy doctors.
Breast screening under pressure
Here in the UK we’re lucky enough to have a highly effective NHS breast screening programme, offering regular mammograms to all women from the age of 50. And research suggests that in England alone the programme saves more than a thousand lives every year.
But it’s also a programme under pressure. The demands of providing three-yearly screenings (thought to be the optimal interval) can prove too much in some areas, and the time period can slip to three and a half or even four years. This is bad news, when you consider that early detection is the key to successful cancer treatment.
But one of the main bottlenecks in the screening system isn’t necessarily the lack of screening staff or units – although we could always do with more of them – but the fact that all mammograms have to be read by not one but two highly-trained specialist radiologists.
Computers to the rescue
To address this, Professor Fiona Gilbert and her colleagues carried out a couple of important studies, funded by Cancer Research UK, to find out whether computers could help out.
The first study, CADET 1 (published in 2006), took more than 10,000 mammograms from 1996 that had been read at the time by two doctors. These mammograms were then read both by a doctor and a computer-aided detection , or CAD, system.
CAD works by scanning mammograms and highlighting ‘dodgy-looking’ areas. These are then scrutinised by a doctor to check if they are likely to be cancer.
The team found that one doctor plus the computer was just as effective at picking up cancers as two doctors.
But this was what’s known as a retrospective study, looking at old mammograms. The researchers were concerned that without the pressure of a real life screening situation, radiologists in the trial would be more likely to over-call, flagging up a higher number mammograms “just in case”.
So the researchers embarked on CADET 2, a prospective “real life” study. More than 28,000 women were recruited from the breast screening programme in Nottingham, Manchester and Coventry, and their mammograms were read by both CAD plus one doctor, as well as the conventional two doctors.
Pleasingly, the results showed that one radiologist plus CAD is just as effective as two when it comes to spotting potential cancers, proving that the system could be viable for the NHS screening service.
It’s unlikely that CAD will ever completely replace the need for trained radiologists – the system still needs a human eye to distinguish potential tumours. highlighted by the computer’s analysis But by effectively halving the workload for busy doctors, the technology could help to ease the pressure on breast screening services.
Click the player below to listen to Professor Gilbert talk to Science Information Officer Alison Ross about CADET.