Artificial Intelligence for the Other Half of Detection at Colonoscopy
Douglas K. Rex, MD, FASGE, reviewing Thakkar S, et al. Gastroenterology 2020 Jan 13.
Assuming that a colonoscopist knows what precancerous lesions look like, high-level detection requires two fundamental steps: (1) exposure of all colonic mucosa and (2) recognition of all precancerous lesions that are exposed. Thus far, artificial intelligence (AI) programs to improve detection have focused on the latter, highlighting lesions when they come into the endoscopic field of view.
The current study is a description of an AI program that is one of a few to address the other half of the issue, assessing the quality of the endoscopist’s mucosal exposure efforts. This paper describes both the development of the program and a proof-of-concept study and provides live case studies that demonstrate the feasibility of the program.
COMMENT
Although this report is mostly qualitative, it demonstrates that AI programs that will assess all aspects of detection will be available in the future. If they become affordable and widely available, they could dramatically reduce the operator dependence of detection at colonoscopy.
Note to readers: At the time we reviewed this paper, its publisher noted that it was not in final form and that subsequent changes might be made.
Douglas K. Rex, MD, FASGE
CITATION(S)
Thakkar S, Carleton NM, Rao B, Syed A. Use of artificial intelligence-based analytics from live colonoscopies to optimize the quality of the colonoscopy exam in real-time: proof of concept. Gastroenterology 2020 Jan 13. (Epub ahead of print) (https://doi.org/10.1053/j.gastro.2019.12.035)
