The radiologists were stumped. They had been staring at a mammogram for several minutes, and couldn’t work out what was wrong.
They sent the scan on to an arbitrator radiologist who declared that it was clear. That’s when Mia, an artificial intelligence (AI) system developed by British start-up Kheiron Medical, stepped in.
The AI immediately highlighted in green an area of concern, a pair of grey clouds which appear to the human eye exactly like the rest of the scan. This section of the scan, Mia concluded, was a sign of cancer.
This could become a familiar scenario across thousands of hospitals in the UK in the coming months as the health service struggles to cope with a backlog of cancer screening and diagnostic appointments due to the pandemic.
“Prior to Covid we had a significant shortage of radiologists within the workforce,” says Dr Jeanette Dickson, the president of the Royal College of Radiologists.
Dickson is concerned that delays in carrying out scans is causing cancers to go undetected. “We know we must be missing some. How many? Who knows,” she says. “That’s the thing we’re worried about.”
Dr Rosie Loftus, the chief medical officer at Macmillan Cancer Support, says there’s fast becoming a “devastating backlog.”
Many health experts are now seeing AI as the best way to fix this. London-headquartered Kheiron Medical is hoping to address the problem by replacing one of the two human radiologists who screen mammograms with its Mia AI system.
Systems like Mia use machine learning, a form of AI, to gradually teach themselves what the signs of cancer are. The AI is fed tens of thousands of scans along with data on whether the women scanned did have cancer. Eventually, it’s hoped the system will be reliable enough to be used in NHS trusts across the country.
There’s certainly an urgent need for help. In August, the Royal College of Radiologists published research based on data from December which found that England requires another 1,600 radiologists, meaning it’s currently 35pc short-staffed. The problem has only gotten worse.
It’s tempting to entirely replace radiologists with AI, solving the backlog problem overnight, but start-ups like Kheiron don’t want to want to remove humans entirely.
“It’s always extremely important for the human to have the authority,” says chief executive Peter Kecskemethy. “Especially in the early phases but probably also longer.”
Kecskemethy, whose mother was a radiologist, likens the technology to the use of electrocardiograms in hospitals. They’re no longer a breakthrough technology, but healthcare professionals still take over if their readings are unclear.