Impact of Artificial Intelligence on Gastroenterology Training and Education: A Survey of Fellows' Perspectives

Introduction: Artificial intelligence (AI) and computer-assisted polyp detection are promising augmentations to endoscopic practice. Clinical trials have linked AI usage to improvements in adenoma detection rates and adenomas detected per colonoscopy. While previous studies have demonstrated generally positive physician sentiments, AI’s effect on gastroenterologist (GI) training is unclear. We aimed to assess GI fellow perspectives on the potential impact of AI on training. Methods: This was a cross-sectional study in which an online survey was sent to 1st, 2nd, and 3rd year GI fellows at a large, urban, academic tertiary care center. Respondents were surveyed on demographics, experience, interest, AI’s potential influence on endoscopy, and attitudes towards implementation. Descriptive statistics were used to summarize their views. Results: A total of 10 invited GI fellows at our institution completed the survey June 2022. Fellows were 80% male and consisted of 3 in their 1st year, 3 in their 2nd year, and 4 in their 3rd year. The majority (60%) had previously used AI platforms. Nine (90%) stated they were interested in using new AI technologies to assist in polyp detection during colonoscopies. There was strong support towards the utilization of AI with 80% of fellows reporting they believe incorporating AI will become the future standard of care and that it would improve detection of adenomas (80%) and sessile serrated lesions (90%). Fellows placed emphasis on validation from AI during colonoscopies; 80% reported they would feel comfortable leaving behind a polyp perceived as hyperplastic by both endoscopic appearance and AI, but only 30% agreed when a polyp is perceived as hyperplastic by appearance but not by AI. Conversely, 80% of fellows felt AI would increase practice costs and procedure times, 90% stated AI is likely to make fellows dependent on technology to discriminate between polyp types, and 90% believe AI will increase false positive polyp detections. Eight (80%) supported the integration of AI technology during fellowship with most recommending integration during the second year. Conclusion: GI fellows show strong interest and optimism towards the utility and potential of AI-assisted polyp detection. However, cost barriers and overreliance leading to training deficits are concerns regarding implementation of AI platforms. Future studies may further elucidate the impact of AI on GI fellowship training and determine optimal time of incorporation, if at all (Figure 1, Table 1).Figure 1.: Fellow survey results regarding AI-assisted colonoscopy. Table 1. - Summary of Fellow Survey Results Regarding AI-assisted Colonoscopy Strongly Agree Somewhat Agree Neither Somewhat Disagree Strongly Disagree I am interested in utilizing new artificial intelligence (AI) technologies that will assist in polyp detection during colonoscopies 4 5 1 0 0 I believe incorporating AI will become the standard of care for polyp detection during colonoscopies in the future 3 5 1 1 0 I believe new AI technologies will generally improve adenoma detection rates 4 4 1 1 0 I believe new AI technologies will generally improve sessile serrated lesion detection rates 6 3 1 0 0 I believe new AI technologies will increase the number of polyps removed per procedure 2 4 4 0 0 I believe AI technologies will make the cost to practice more expensive 5 3 1 1 0 AI will decrease my total time dedicated per procedure 0 1 1 7 1 I believe AI technologies are worth the investment to improve endoscopist performance 1 6 1 2 0 I believe endoscopists are currently able to reliably differentiate between hyperplastic and adenomatous polyps based on endoscopic appearance without AI assistance 0 9 0 1 0 I would feel comfortable leaving behind a polyp I perceive as hyperplastic from endoscopic appearance alone 1 9 0 0 0 I would feel comfortable leaving behind a polyp I perceive as hyperplastic from endoscopic appearance with validation from an AI platform 5 3 2 0 0 I would feel comfortable leaving behind a polyp I perceive as hyperplastic from endoscopic appearance but an AI platform does not 0 3 2 3 2 AI will make fellows dependent on technology to detect polyps 1 6 1 2 0 AI will increase my comfort and confidence in not missing polyps during colonoscopies 1 5 3 1 0 AI will detect lesions I otherwise would not have 2 5 2 1 0 AI will increase false positive polyp detections during colonoscopies 5 4 1 0 0 Access to AI during training would make me more likely to choose a fellowship program over another without such access 1 4 1 4 0 Yes No AI will be beneficial for fellows to incorporate during training 8 2 1st Year 2nd Year 3rd Year If you believe AI will be beneficial for fellows to incorporate during training, when should AI be optimally integrated during fellowship? 0 5 3