A Northwestern Medicine study of nearly 18,600 colonoscopies performed by 55 physicians found that an AI tool could accurately assess several colonoscopy quality metrics from procedure videos. Its measurement of withdrawal time closely matched nurse-recorded times, while the system also tracked metrics such as polyps removed and use of cold snare polypectomy—measures the researchers say are difficult to assess manually at scale.
The tool could provide a scalable way for hospitals and health systems to routinely monitor colonoscopy performance and give clinicians feedback. The researchers emphasize that it is designed to assess procedures after they are completed, rather than replace the endoscopist. The study also raises an important question about AI and clinical skills: whether AI-driven feedback can identify blind spots and improve performance or, conversely, contribute to physician deskilling.