A new feature by Prateek Sharma, MD, and Jillian Mock examines two developments in endoscopy: successful repeat colonoscopy after an incomplete examination and the creation of a structured video databank for AI research.
Repeat colonoscopy: A 24-year retrospective study of 1,334 patients at Indiana University found that repeat colonoscopy achieved 96.1% cecal intubation overall after an unsuccessful initial examination. Success rates were 97.0% for difficult sigmoid colons and 96.9% for redundant colons. Of 31 lesions measuring at least 20 mm detected during repeat colonoscopy, 20 had not been identified in prior radiographic studies. The findings support referral to experienced endoscopists before assuming an incomplete examination cannot be completed.
Building an endoscopy AI databank: Researchers at the University of Montreal Hospital Research Center developed a prospective database containing 10,831 procedures from 8,658 patients. Full-length endoscopy videos are linked to structured procedural annotations, lesion characteristics and pathology findings to support the development and evaluation of clinically relevant AI tools.
Why it matters: The first study informs decisions after incomplete colonoscopy; the second highlights the data infrastructure needed to develop and validate endoscopy AI.