S1248 Training Future Endoscopists: Gastroenterology Fellows’ Perspectives and Hands-On Exposure to Artificial Intelligence for Polyp Detection in the United States

Introduction: Artificial intelligence (AI)-assisted colonoscopy for polyp detection is a promising tool to improve colonoscopy quality. While surveys have assessed staff gastroenterologists’ attitudes towards AI, less is known about the views of GI fellows regarding AI during training. This study explored GI fellows' exposure to AI, their interactions with it, and their attitudes towards its incorporation into training. Methods: We conducted a nationwide survey of GI fellows from 8/2024-11/2024 to assess: 1) exposure to and experience with AI in fellowship, 2) perceptions of AI’s impact on colonoscopy quality, and 3) attitudes towards implementing AI into training. The survey included Likert-scale questions with branching logic to tailor questions based on AI availability at the fellows’ institutions. Participation was voluntary for all questions. Results: A total of 126 fellows started the survey, and 88 (69.8%) completed it. AI was available at least at 1 training site for 69.3% of respondents, with the highest availability at Veteran Affairs (VA) hospitals. Most fellows (72.4%) believed AI improved adenoma detection, and 88.6% felt it reduced adenoma miss rates. They also noted potential drawbacks, such as increased benign lesion resection rates and longer withdrawal times due to double-checking false positives. Notably, 66% of fellows reported visualizing a polyp before AI detection, 55.7% stated it occurred often. 81.8% of fellows felt AI should be available during fellowship. Many fellows (43.2%) thought AI should be incorporated in the 2nd year of training, compared to 31.8% in the 1st year, 18.2% in the 3rd year, while 6.8% disagreed with its inclusion in training. Most fellows (60.7%) believed early exposure to AI enhanced polyp detection skills. However, 52.5% felt neutral that AI made them better endoscopists overall. Despite this, 62.5% preferred to pursue a job with AI if they had trained with it. Conclusion: Our nationwide survey found that GI fellows are generally supportive of integrating AI into their training, with many advocating for its incorporation in the 2nd or 3rd year (61%). These results are important to consider for fellowship program leadership and GI practices recruiting fellows trained with AI. Further studies are required to assess the impact of training GI fellows with AI on their polyp detection competency.