ASGE International Sampler (On-Demand)-Clinical Evaluation of ML Algorithms in Gastroenterology

Pdf Summary

The document discusses the clinical evaluation of machine learning algorithms in the field of gastroenterology, focusing on computer-aided polyp detection (CADe) and diagnosis (CADx). It highlights the need for evidence to adopt CADx and practical considerations for evaluation in endoscopy suites. Studies have shown that AI consistently improves adenoma detection rates, particularly for adenomas 5mm in size. The use of AI during colonoscopy has been shown to reduce adenoma miss rates and increase the need for intensive surveillance. The implementation of CADe in real practice is being studied through observational research, which has shown some concerns regarding endoscopist performance. Additionally, the document emphasizes the importance of CADx in encouraging AI adoption for colonoscopy, with the potential for cost-savings. It also mentions the challenges in critically evaluating AI tools in GI units and the importance of measuring relevant metrics. The integration of AI with human decision-making can lead to better outcomes in clinical practice. The document concludes by stressing the significance of NLP-powered analytics in gastroenterology practices and the need for further research and development in the field of AI and ML for improved patient outcomes.

Asset Subtitle

Tyler Berzin MD, MS, FASGE

Meta Tag

Concept Computer-Aided Polyp Detection

Concept Lower Endoscopy

Concept Computer-aided Diagnosis

Concept Adenoma Detection Rate

Concept Polyp Miss Rate

Keywords

machine learning

gastroenterology

computer-aided polyp detection

computer-aided diagnosis

adenoma detection

colonoscopy

AI in endoscopy

endoscopist performance

NLP-powered analytics

clinical outcomes

Computer-Aided Polyp Detection

Lower Endoscopy

Computer-aided Diagnosis

Adenoma Detection Rate

Polyp Miss Rate