Radboudumc is seeing promising results in the use of artificial intelligence (AI) during coloscopy, a keyhole examination of the colon using an endoscope. The technology yields 40 per cent more detected polyps, which are structures that can eventually develop into cancer.
This is according to a publication by Radboudumc in *The Lancet Digital Health*. More than 100,000 Dutch people undergo a colon examination, also known as a colonoscopy, every year. During the procedure, a doctor uses a camera to examine the bowel, looking for bowel cancer and polyps. Polyps are growths on the lining of the bowel and can be a precursor to bowel cancer.
More polyps removed leads to lower risk of colon cancer
Previous research already indicated that the more polyps the doctor finds and removes during a colonoscopy, the less likely a person is to develop colon cancer later. The research results now presented by Radboudumc show that AI can add great value here. For example, the technology appears to increase the number of polyps detected by 40%. This therefore has a direct impact on the chance of someone later developing bowel cancer.
A total of nearly 1,000 patients participated in the Radboudumc study. It was conducted in 10 hospitals in several countries. Participants in the study were all eligible for colon cancer screening, or have had polyps before. Patients were divided into two groups. One group received a standard colonoscopy, while in the second group the AI watched live.
Purple square attracts attention
The AI places a purple square in the image at each suspicious spot. That square is meant to attract the attention of the doctor, who then looks more closely at the suspicious image. The method led to 40% more polyps detected.
Physician-researcher Michiel Maas speaks of enormous added value. "A doctor may miss polyps because they are just too short in the image, or the attention is focused on a different spot," says Maas. "But a computer looks at every pixel in the image and is equally alert every millisecond. That adds enormous value."
Examination does not take more time
The researchers point out that although the number of polyps detected and removed is higher, the use of AI does not increase the time taken for coloscopy in practice.
Also, the number of false positives has not increased. Examination of the tissue removed shows that the extra polyps removed are as often of a dangerous type as those excised during a standard colonoscopy.
Developed by Magentiq Eye
The AI used by Radboudumc researchers was developed by Israel-based company Magentiq Eye. This company has been operating since 2014. It focuses on developing AI for AI-assisted colonoscopy, in which a doctor examines the mucosa of the colon.
Magentiq Eye's system - MAGENTIQ-COLO - is compatible with commonly used endoscopy systems, including those from Olympus, Fujinon, Pentax and Sonoscape. The system uses a digital video signal from the endoscopy system, and applies a deep-learning algorithm to it in real time. In doing so, the AI looks for abnormalities on the images that may indicate a polyp.
These suspicious images are brought to the doctor's attention by Magentiq Eye. This can be displayed on the main screen of the system if required, or on a separate screen. It not only displays the suspicious spot, but also shows an estimate of the size and characteristics of the polyp in the image, among other things. Thus, during the examination, the system provides additional information to the doctor, who can use it for his analysis. At the end of the procedure, the recorded video can be exported and saved with or without marked suspicious spots. Finally, MAGENTIQ-COLO automatically produces a report on the findings.
Deployment still limited
MAGENTIQ-COLO is commercially available, but its use is still limited. Maas: "This system costs money, and reimbursement for medical research using AI is still in the development phase."
Follow-up studies are planned. Among other things, the researchers aim to demonstrate the extent to which MAGENTIQ-COLO is able to determine whether a polyp is dangerous based on external features and size. Professor and MDL physician Peter Siersema: "If we can determine that during the study, we can leave innocent polyps in place. So there is no need to have them analysed at Pathology. Ultimately, that saves on the risk of complications and on costs."
Author: Wouter Hoeffnagel
Image: Julio César Velásquez Mejía via Pixabay
