Gastroenterology & Artificial Intelligence: 3rd Annual Global Artificial Intelligence Summit | September 2021-AI and Radiology: Lessons for Clinical Implementation in GI

AI and Radiology: Lessons for Clinical Implementation in GI

Video Transcription

Video Summary

In this video, Dr. Chuck Kahn, professor and vice chair of radiology at UPenn, discusses the lessons learned from applying artificial intelligence (AI) in radiology and how they can be applicable to gastroenterologists and endoscopy. He emphasizes the importance of rigorous testing and understanding what AI systems have learned. Dr. Kahn presents examples of systems that mistakenly focused on non-relevant features, such as the letter "L" instead of pneumonia, or detected unrelated objects like orthopedic hardware instead of fractures. He explains the challenges of measuring the performance of AI systems, including the Dice Similarity Coefficient and calibration. Dr. Kahn also highlights the significance of defining ground truth and the need for standardized terminology and data elements. He encourages the use of AI competitions and challenges to advance the field. Dr. Kahn concludes by urging practitioners to test rigorously, understand ground truth, develop standards, and participate in AI competitions.

Asset Subtitle

Charles Kahn, MD

Keywords

artificial intelligence

radiology

gastroenterologists

rigorous testing

ground truth