Gastroenterology and Artificial Intelligence: 2nd Annual Artificial Intelligence Summit | October 2020-Winning on Detection, Classification and Navigation Using AI

Winning on Detection, Classification and Navigation Using AI - Is it Enough?

Video Transcription

Video Summary

In this video, presented by Ehud Rivlin, the topic of discussion is whether winning on detection, classification, and navigation using AI is enough in medical imaging. Dr. Rivlin starts by highlighting the importance of these three aspects and their relevance to medical errors, which are the third leading cause of death in the US. He focuses on colonoscopy and the sources of error in detecting polyps, including missed targets, incomplete coverage, and misclassification. He states that detection is almost at a satisfactory level with the availability of computer-aided detection, but classification and navigation still need improvement. Dr. Rivlin then explains the concept of colon coverage and proposes a method for its computation using depth extraction and trajectory reconstruction. He discusses the unsupervised learning approach used to estimate both depth and pose, and showcases results on real data and evaluation by gastroenterologists. In conclusion, Dr. Rivlin suggests that winning on the detection, classification, and navigation trio is not enough, and there is a need for personalized, comparative, and decision-making capabilities, as well as the implementation of pocket technology in endoscopy.

Asset Subtitle

Ehud Rivlin, PhD, MSc

Keywords

medical imaging

detection

classification

navigation