Doctors at Sengkang General Hospital (SKH) are using a newly developed artificial intelligence (AI) screening tool that serves as a second pair of eyes to spot near-invisible cancerous growths in a patient’s organs.
Developed with the help of biomedical company Medtronic, the tool is able to detect roughly 20 per cent more growths – or polyps – that doctors would otherwise miss with the human eye, according to studies by SKH.
The detection rate is expected to increase as the AI learns to recognise more polyps.
The screening tool, which was adopted at the hospital in July 2021, will be available in all its colonoscopy suites by mid-2023, SKH said on Tuesday during the signing of an agreement to establish a Centre of Excellence for AI-assisted Colonoscopy, which will launch further studies and training in the field.
SKH is among the early adopters of AI for detecting cancer in hospitals and is the first among public health institutions here.
Colorectal cancer is Singapore’s leading source of cancer, affecting more than 1,800 people each year, especially among those above 50 years old, according to the National University Cancer Institute, Singapore.
But the cancer is also known to affect those younger – such as Black Panther star Chadwick Boseman, who died at 43 in 2020.
The cancer usually starts as a non-cancerous polyp on the inner lining of the colon or rectum, which may develop into cancer years later.
Early detection through colonoscopy is key to preventing the cancer, said SKH senior consultant Winson Tan, 40, adding that the development of AI in this field will speed up the detection of polyps in patients to help them act fast.
Dr Tan, who specialises in colorectal surgery, said: “Some polyps are hard to find because they can be very subtle and appear almost normal.
“The AI also makes handling of numerous colonoscopy procedures in a day more manageable as it uses pattern recognition that spots polyps quickly and may even detect polyps that a doctor might miss due to fatigue.”
The polyp detection module can be plugged into most major brands of endoscopic equipment and helps to analyse in real time footage that is gathered from a pinhole camera during surgery.
The AI is trained on a growing pool of at least 13 million images taken during colonoscopy treatment to recognise growths of all sizes, even those that appear flat and nearly invisible to the human eye.
In a hands-on session for the media in SKH, I wore a virtual reality (VR) headset to experience how doctors conduct colonoscopy and how they can use the AI tool to seek out polyps.
With VR controllers in hand, I picked up the digital endoscope that was attached to a virtual patient to begin the demo.
I was asked to analyse images of the patient’s insides and tap the areas where I think a polyp might lie. I failed to get any of my three attempts right as the lumps were subtler than they seem.
A year-long study by SKH found that the AI was able to find polyps in 34 per cent of 2,433 colonoscopy patients, regardless of whether they were cancerous. Meanwhile, doctors found polyps in only one in four of 1,770 colonoscopies without the help of AI.
The improvement in detection rates is significant, said Dr Tan, who has been practising for 14 years. “Every percentage increase counts because the more pre-cancerous polyps we detect and remove, the more cases of cancer we can help to prevent,” he said.
SKH is among several hospitals that have worked with AI in cancer treatment, including Farrer Park Hospital, which has adopted a screening service to spot polyps.
At Tan Tock Seng Hospital, the ability of AI in blood tests to diagnose leukaemia and malaria is also being tested.
Radiologists in Hungary are also testing how AI can be used to help detect breast cancer.