6th Global Gastroenterology and Artificial Intelligence Summit (Live) | September 2024-Evaluating the Accuracy and Safety of AI-Driven Endoscopy

Pdf Summary

The document authored by Dr. Cesare Hassan discusses the utilization of Artificial Intelligence (AI) in the field of endoscopy, specifically in gastroenterology. Key points cover AI's integration through computer vision, Natural Language Processing (NLP), large language models (LLMs), and multi-modal LLMs, all of which aid in enhancing the diagnostic capabilities of endoscopic procedures.<br /><br />The document extensively details the diagnostic performance of AI in detecting colorectal polyps, showcasing sensitivity rates of 95% and specificity rates of 88%. AI's primary applications in endoscopy include Computer Aided Detection (CADe) for identifying polyps and Computer Aided Diagnosis (CADx) for characterizing these polyps. The benefits of using CADe are significant, highlighted by improved adenoma detection rates and reduced adenoma miss rates compared to traditional methods without AI assistance.<br /><br />Dr. Hassan's analysis incorporates multiple clinical studies and randomized controlled trials, confirming that AI-assisted endoscopy markedly increases the adenoma detection rate (44%) compared to control groups (37.9%), and lowers adenoma miss rates (16% for CADe-assisted versus 35% for the control group). This collective evidence is derived from a meta-analysis of 21 randomized controlled trials involving over 18,000 patients.<br /><br />Despite AI's high accuracy, it is stressed that human expertise (GI-competence) remains superior, particularly in critical tasks such as correcting AI errors, distinguishing between true positives (AI-TP) and false positives (AI-FP), and ensuring optimal clinical outcomes. The document also notes the specific benefits and potential harms of using AI for polyp characterization and underscores the need for a balanced approach integrating both AI and human expertise to maximize diagnostic accuracy and patient safety.<br /><br />In summary, AI significantly contributes to advancements in endoscopy by enhancing polyp detection and characterization. However, the successful and safe implementation of AI technologies depends on the combined expertise of AI systems and trained gastrointestinal specialists.

Asset Subtitle

Cesare Hassan, MD, PhD

Meta Tag

Concept Computer-Aided Detection

Concept Computer-aided Diagnosis

Concept Lower Endoscopy

Concept Computer Vision

Concept Large Language Model

Keywords

Artificial Intelligence

endoscopy

gastroenterology

computer vision

Natural Language Processing

large language models

Computer Aided Detection

Computer Aided Diagnosis

adenoma detection rates

clinical studies

Computer-Aided Detection

Computer-aided Diagnosis

Lower Endoscopy

Computer Vision

Large Language Model