Robot scans that can spot bowel cancer humans miss

Artificial intelligence may be more effective than the human eye alone at spotting the early signs of bowel cancer.

A new UK trial is investigating whether adding AI technology — which uses computer algorithms to scan and read images — to standard colonoscopy examinations improves the accuracy of these scans.

More than 42,000 people are diagnosed with bowel cancer in the UK each year and 16,000 die from it, making it the second most common cause of cancer death.

Colonoscopies are the ‘gold standard’ way of diagnosing the disease. This is where the large bowel is examined using a camera attached to a thin, flexible tube.

The camera relays live images from inside the bowel on to a screen, allowing the clinician carrying out the procedure to check for pre-cancerous polyps called adenomas — small growths that can be found on the wall of the bowel. It is believed that bowel cancer develops from these polyps and, if detected, they can be removed during the procedure.

However, although colonoscopies are extremely effective, three in every 100 examinations miss a cancer or polyp which might be small, flat or hidden in the folds of the bowel but which goes on to become a cancer, according to the NHS.

Scientists hope that integrating AI technology (which not only reads scans but also learns as it goes along) into existing colonoscopy equipment could help save more lives by boosting the accuracy of the 45-minute procedure, so that more cancers are caught at an early stage when they’re easier to treat.

To try to locate these hard-to-find abnormalities, U.S. researchers have developed an AI box called the GI Genius which connects to colonoscopy equipment and analyses the video footage in real time.

If it spots something unusual, the device creates a green box on the screen pinpointing a precise section of the bowel lining which needs closer inspection and sounds an alert. The medic carrying out the scan will then decide whether to investigate further.

The first UK trial to test the AI device is halfway through screening around 2,000 NHS patients.

Patients enrolled on the trial have either undergone a colonoscopy before or have experienced symptoms such as blood in their stools or significant changes to their bowel habits, which they have reported to their GP; or have taken part in the NHS Bowel Screening Programme (a home testing kit sent to adults aged 60 to 74 in England, and from the age of 50 in Scotland).

Half of those on the trial will undergo a standard colonoscopy, the other half with the AI device.

Nine hospitals across England are taking part in the Colo-Detect trial, mostly in the North East, with the study being led by Newcastle University and South Tyneside and Sunderland NHS Foundation Trust.

The trial, funded by U.S. medical device company Medtronic that designed the device, is due to end in April (researchers will evaluate both the clinical and cost-effectiveness of the technology).

Results of the first U.S. trial of the device, published in the American Journal Gastroenterology last year, showed a 50 per cent reduction in missed polyps when the AI technology was used compared to standard colonoscopy.

Commenting on the new trial, Dr Duncan Gilbert, a consultant clinical oncologist in lower gastrointestinal cancers at University Hospitals Sussex NHS Foundation Trust, said: ‘Colorectal cancer remains a major public health challenge for the UK. Worryingly, it is also becoming more common in younger patients.

‘Screening using colonoscopy to find and remove polyps and early cancers has been shown to save lives and anything that improves the effectiveness of colonoscopy is to be welcomed.

‘Testing new technologies in properly conducted clinical trials such as this is exactly what we need to do and is an example of how NHS clinical research leads the world.’