Colonoscopies are one of the most effective ways to prevent colorectal cancer, which is surging among young adults in the U.S. Yet, the quality of each procedure can vary substantially from one physician to another. As a result, medical societies recommend routinely reviewing every colonoscopist - something that's challenging and time-consuming for practices to implement.
Now, a team of Northwestern Medicine scientists has developed an AI tool that rapidly and successfully measures colonoscopy quality by reviewing footage of the procedure. See a video demonstrating how the tool works here.
In a study of nearly 19,000 colonoscopies published last month in The American Journal of Gastroenterology, the Northwestern team demonstrated the tool's high accuracy by comparing its quality measurements with those of nurses and other clinicians. The authors say the study marks the first time an AI tool is shown to comprehensively measure the quality of thousands of colonoscopies.
"Our AI software provides a scalable approach to measuring colonoscopy quality and, ultimately, providing feedback to clinicians to improve care quality," said study lead author Dr. Rajesh Keswani, associate professor of medicine in the division of gastroenterology and hepatology at Northwestern University Feinberg School of Medicine.
Colonoscopies let physicians examine patients' large intestines, or colons, via an inserted tube that's equipped with a camera and with tools to remove precancerous polyps. A high-quality colonoscopy generally means a physician inspects the entire colon, spends enough time looking for abnormalities and uses recommended techniques to remove suspicious polyps.
Previous studies have shown that measuring colonoscopy performance improves the quality of colonoscopy overall and, ultimately, reduces colorectal cancer mortality. By automating that quality-monitoring process, the new tool could make colonoscopy feedback more feasible across entire hospitals or healthcare systems, said Keswani, who also is a Northwestern Medicine physician.
"The goals of screening colonoscopy are to provide safe and effective care," he said. "This is not possible unless an institution can measure the quality of colonoscopy."
Findings
The AI tool analyzed nearly 18,600 colonoscopies performed by 55 physicians over 11 months at Northwestern. For the study, the AI software reviewed colonoscopy recordings and identified key moments, such as when the tube reached the beginning of the colon, when it was withdrawn and when polyps were removed.
The tool's measurement of withdrawal time - a key metric that tracks time spent examining the colon as the scope is being withdrawn - closely matched times recorded by nurses. That showed the technology is highly accurate, the authors said.
The AI also tracked quality indicators that "can't be feasibly measured by humans at scale," according to Keswani. These included the number of polyps removed during a procedure and how often physicians used cold snare polypectomy, a guideline-recommended technique for removing small polyps.
AI concerns
Last year, a Lancet study created headlines after it suggested that colonoscopists who started using an AI helper to detect polyps were becoming less proficient over time. Keswani said he's not sure yet how to view the impact of AI on colonoscopies and on medicine in general. He also noted that his AI tool only assesses quality after colonoscopies have been completed.
"One possibility is that physicians rely too much on AI, which leads to deskilling," he said. "Alternatively, AI can teach us about blind spots and make us better clinicians."
Keswani added that his research team is currently studying the role and impact of AI in teaching colonoscopy to trainees.
Other Northwestern authors are Dr. John Pandolfino, Dr. Mozziyar Etemadi, Matthew Wittbrodt, Alex Heller, Kristjana Kristinsdottir and Evandros Kaklamanos.
The study is titled, "Artificial Intelligence Automated Assessment of Colonoscopy Quality Metrics."
