Gastroenterology & Artificial Intelligence: 3rd Annual Global Artificial Intelligence Summit | September 2021-Panel Discussion and Q&A

Video Transcription

Video Summary

In this video discussion, Dr. Khan and Dr. Hassan emphasize the importance of training AI systems with well-annotated and carefully curated datasets. They highlight that having high-quality datasets can significantly improve the performance of AI systems in medical imaging. Dr. Khan also mentions a publicly available dataset from the NIH called the chest X-ray 14 dataset, which was annotated by a team of thoracic and neuroradiologists. They discuss the challenges of training AI systems in endoscopy, particularly in terms of having a diverse and representative dataset. Dr. Michael Riegler, Chief Research Scientist at Simulanet, shares his experience with creating open datasets for endoscopy and the importance of sharing and collaborating globally to build useful AI systems. Dr. Shani Haugen, Assistant Director for the Gas Entomology and Endoscopy Devices team at the FDA, discusses the need for standardizing terminology and information presented to physicians in the labeling of AI algorithms. She also mentions the importance of explaining how AI algorithms work to clinicians and patients. Overall, the discussion highlights the challenges and considerations in training and evaluating AI systems in the field of medical imaging and endoscopy.

Keywords

training AI systems

high-quality datasets

medical imaging

endoscopy

open datasets

standardizing terminology