Post written by Daniela Guerrero Vinsard, MD, and Yuichi Mori, MD, PhD, from the Showa University International Center for Endoscopy and Digestive Disease Center, Showa University Northern Yokohama Hospital, Yokohama, Japan, and the Division of Internal Medicine, University of Connecticut Health Center, Farmington, Connecticut, USA.
Artificial Intelligence (AI) in gastrointestinal endoscopy entails the construction of an intelligent software which, once installed in the endoscopic unit, has the potential of enhancing the endoscopist’s adenoma detection and adenoma characterization, with possible reduction of colorectal cancer rates. The focus of this study is to educate the readers on current AI advantages and disadvantages and propose the ideal preclinical manufacture of the software, clinical software application, and post-clinical impact measures to ensure the highest quality for development of this technology.
The purpose of this review is to propose some principles for system development, clinical testing, and implementation of AI in colonoscopy, in anticipation of future research involving this novel technology. The vast majority of AI systems have been evaluated in an experimental or retrospective fashion and have already shown promising results for adenoma detection and characterization. Moving forward, in-vivo randomized controlled trials are the ideal methodology to analyze the tool’s performance, safety, and limitations; our publication acts as a guide for researchers interested in this field.
Previous publications have rarely addressed the downsides of AI such as extra time required for colonoscopy, the risk of endoscopist distraction with the output from CADe/CADx, and “over-reliance” on AI which may make the new generation of endoscopists less skillful given the sense of security provided by this tool. We highlight in this publication that AI tools should by no means act as an autonomous decision maker. Its application should be targeted to act as an enhancer of the endoscopists’ ability to detect and characterize adenomas. We also believe that early recognition of advantages and disadvantages of the use of AI could facilitate the development of high-quality AI systems.
Figure 1. Two types of outputs for automated polyp detection. A, Presence of the polyp is indicated by a visible alarm outputting color outside the endoscopic monitor. B, Polyp location is indicated by a visible rectangle.
There is an increasing number of gastroenterologists and bio-engineering partners interested in the development of this technology. To secure a successful and appropriate introduction of this innovation, we must seek for the highest quality in publications reporting the development and validation of AI systems in a standardized manner.
Read the full article online.
The information presented in Endoscopedia reflects the opinions of the authors and does not represent the position of the American Society for Gastrointestinal Endoscopy (ASGE). ASGE expressly disclaims any warranties or guarantees, expressed or implied, and is not liable for damages of any kind in connection with the material, information, or procedures set forth.