Jiyoung Lee1,2, Michael B Wallace3,4. 1. Division of Gastroenterology and Hepatology, Endoscopy Unit, Mayo Clinic Jacksonville, 4500 San Pablo Road, Jacksonville, FL, 32224, USA. 2. Health Screening and Promotion Center, Asan Medical Center, Seoul, South Korea. 3. Division of Gastroenterology and Hepatology, Endoscopy Unit, Mayo Clinic Jacksonville, 4500 San Pablo Road, Jacksonville, FL, 32224, USA. wallace.michael@mayo.edu. 4. Center of Research in Computer Vision, University of Central Florida, Orlando, FL, USA. wallace.michael@mayo.edu.
Abstract
PURPOSE OF REVIEW: Recently numerous researchers have shown remarkable progress using convolutional neural network-based artificial intelligence (AI) for endoscopy. In this manuscript we aim to summarize recent AI impact on endoscopy. RECENT FINDINGS: AI for detecting colon polyps has been the most promising area for application of AI in endoscopy. Recent prospective randomized studies showed that AI assisted colonoscopy increased adenoma detection rate and the mean number of adenomas per patient compared to standard colonoscopy alone. AI for optical biopsy of colon polyp showed a negative predictive value of ≥90%. For capsule endoscopy, applying AI to pre-read the video images decreased physician reading time significantly. Recently, researchers are broadening the area of AI to quality assessment of endoscopy such as bowel preparation and automated report generation. AI systems have shown great potential to increase physician performance by enhancing detection, reducing procedure time, and providing real-time feedback of endoscopy quality. To build a generally applicable AI, we need further investigations in real world settings and also integration of AI tools into pragmatic platforms.
PURPOSE OF REVIEW: Recently numerous researchers have shown remarkable progress using convolutional neural network-based artificial intelligence (AI) for endoscopy. In this manuscript we aim to summarize recent AI impact on endoscopy. RECENT FINDINGS: AI for detecting colon polyps has been the most promising area for application of AI in endoscopy. Recent prospective randomized studies showed that AI assisted colonoscopy increased adenoma detection rate and the mean number of adenomas per patient compared to standard colonoscopy alone. AI for optical biopsy of colon polyp showed a negative predictive value of ≥90%. For capsule endoscopy, applying AI to pre-read the video images decreased physician reading time significantly. Recently, researchers are broadening the area of AI to quality assessment of endoscopy such as bowel preparation and automated report generation. AI systems have shown great potential to increase physician performance by enhancing detection, reducing procedure time, and providing real-time feedback of endoscopy quality. To build a generally applicable AI, we need further investigations in real world settings and also integration of AI tools into pragmatic platforms.
Authors: Cesare Hassan; Michael B Wallace; Prateek Sharma; Roberta Maselli; Vincenzo Craviotto; Marco Spadaccini; Alessandro Repici Journal: Gut Date: 2019-10-15 Impact factor: 23.059
Authors: Peter Klare; Christoph Sander; Martin Prinzen; Bernhard Haller; Sebastian Nowack; Mohamed Abdelhafez; Alexander Poszler; Hayley Brown; Dirk Wilhelm; Roland M Schmid; Stefan von Delius; Thomas Wittenberg Journal: Gastrointest Endosc Date: 2018-10-17 Impact factor: 9.427
Authors: Douglas A Corley; Christopher D Jensen; Amy R Marks; Wei K Zhao; Jeffrey K Lee; Chyke A Doubeni; Ann G Zauber; Jolanda de Boer; Bruce H Fireman; Joanne E Schottinger; Virginia P Quinn; Nirupa R Ghai; Theodore R Levin; Charles P Quesenberry Journal: N Engl J Med Date: 2014-04-03 Impact factor: 91.245