Northwestern AI tool audits colonoscopy quality at scale

Northwestern Medicine scientists have developed an AI tool that reviews colonoscopy footage to assess procedure quality, aiming to make routine performance monitoring more practical across hospitals and healthcare systems. Medical societies recommend reviewing colonoscopists because procedure quality can vary, but manual review is challenging and time-consuming.

In a study of nearly 19,000 colonoscopies published in The American Journal of Gastroenterology, the team compared the software’s measurements with assessments from nurses and other clinicians and reported high accuracy. The AI reviewed nearly 18,600 colonoscopies performed by 55 physicians over 11 months at Northwestern, identifying key moments such as when the scope reached the beginning of the colon, when it was withdrawn and when polyps were removed.

The system closely matched nurse-recorded withdrawal time, a key measure of how long clinicians spend inspecting the colon as the scope is withdrawn. It also tracked quality indicators that are difficult to measure at scale, including the number of polyps removed and use of cold snare polypectomy, a recommended technique for removing small polyps.

Study lead author Dr. Rajesh Keswani said the tool is designed to assess quality after procedures are completed and could help clinicians improve care. He also noted broader questions about AI in medicine, including whether AI could lead to overreliance or help identify clinical blind spots.