Skip Navigation
Skip to contents

Ann Coloproctol : Annals of Coloproctology

OPEN ACCESS
SEARCH
Search

Articles

Page Path
HOME > Ann Coloproctol > Volume 42(4); 2026 > Article
Original Article
Colonoscopy
Endoscopists’ attitudes toward and experiences with artificial intelligence–assisted colonoscopy: a 3-year follow-up survey after the implementation of GI Genius
Ronja M. B. Lagström1orcid, SKH Endoscopy Centre2,3, Artificial Intelligence in Gastrointestinal Endoscopy Singapore (AIGES)3, Mustafa Bulut1,4,5orcid, Ismail Gögenur1,4orcid, Frederick H. Koh6orcid
Annals of Coloproctology 2026;42(4):436-445.
DOI: https://doi.org/10.3393/ac.2026.00178.0025
Published online: August 25, 2026

1Center for Surgical Science, Department of Surgery, Zealand University Hospital, Køge, Denmark

2Department of Gastroenterology and Hepatology, Sengkang General Hospital, SingHealth Services, Singapore

3Full list of members is available from Supplementary Material 1

4Department of Clinical Medicine, University of Copenhagen, Copenhagen, Denmark

5Copenhagen Academy for Medical Education and Simulation (CAMES), Copenhagen, Denmark

6Department of General Surgery, Sengkang General Hospital, Singapore

Correspondence to: Ronja M. B. Lagström, MD Center for Surgical Science, Department of Surgery, Zealand University Hospital, Lykkebækvej 1, Køge 4600, Denmark Email: rlag@regionsjaelland.dk
• Received: January 25, 2026   • Revised: May 4, 2026   • Accepted: May 9, 2026

© 2026 The Korean Society of Coloproctology

This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/licenses/by-nc/4.0/) which permits unrestricted non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited.

prev next
  • 1,907 Views
  • 31 Download
  • Purpose
    Artificial intelligence (AI) can improve colonoscopy quality by increasing adenoma detection rates. At Sengkang General Hospital (SKH), Singapore, AI-assisted colonoscopy was introduced in July 2021. An initial survey suggested that attitudes toward AI were shaped more by hands-on experience than by general knowledge of AI. This follow-up survey assessed how these attitudes evolved over time and compared responses from SKH with those from endoscopists at institutions without routine AI use.
  • Methods
    An online survey was conducted 3 years after implementation of routine AI-assisted colonoscopy at SKH to evaluate endoscopists’ knowledge, perceptions, and behaviors regarding AI. Endoscopists from 7 tertiary institutions in Singapore were invited to participate. The 44-item questionnaire covered 4 thematic domains. Responses were compared with those from SKH’s initial survey and across institutions.
  • Results
    A total of 57 endoscopists, including 46 consultants and 11 residents, completed the survey. Most respondents were familiar with AI, but only approximately half understood its underlying mechanisms. Most endoscopists supported early AI training, although agreement was significantly lower at SKH than at other hospitals (61.9% vs. 88.9%, P=0.02). Overall, 91.2% agreed that AI should be used as an adjunct to colonoscopy, and 80.7% agreed that AI can improve the quality of a performed colonoscopy. Enjoyment of AI-assisted colonoscopy was higher at SKH than at other hospitals (85.7% vs. 58.3%, P=0.04).
  • Conclusion
    Endoscopists generally perceived AI as improving colonoscopy quality. Greater colonoscopy experience and more extensive exposure to AI-assisted colonoscopy appeared to be associated with more positive attitudes toward AI and greater acceptance of its potential benefits.
Artificial intelligence (AI) can reduce performance variability during colonoscopy by helping compensate for perceptual inconsistencies related to fatigue, distraction, and limitations in human visual recognition [1, 2]. Randomized controlled trials have shown that AI-assisted colonoscopy is associated with significantly higher adenoma detection rates (ADRs) and adenomas per colonoscopy than colonoscopy performed without AI assistance [3–11]. AI-assisted colonoscopy was implemented at Sengkang General Hospital (SKH), Singapore, in July 2021. In an earlier review, GI Genius (Medtronic) improved the ADR in 72.2% of consultant endoscopists who performed at least 5 AI-assisted colonoscopies during the first 3 months after implementation [12]. One year later, this improvement in ADR was maintained, suggesting that adaptation to the AI system did not diminish the increased detection rates [13].
Despite evidence that AI increases ADR, endoscopists’ attitudes toward AI vary. Therefore, an initial survey was conducted 2 months after implementation of GI Genius at SKH to assess endoscopists’ knowledge of AI, perceptions of AI in medicine, and behaviors related to AI-assisted colonoscopy. A total of 24 consultant endoscopists were invited to complete the survey, and 16 responded. The survey suggested that acceptance of AI-assisted colonoscopy was more strongly related to endoscopists’ experience using the AI system than to their general knowledge of AI. Although 68.8% of respondents agreed that GI Genius should be used during colonoscopy, only slightly more than half reported in-depth knowledge of AI [14].
This study evaluated how attitudes toward AI-assisted colonoscopy evolved over time among endoscopists with increasing experience using an AI system. Specifically, responses from a follow-up survey conducted 3 years after the initial survey were compared with those from the initial survey to assess changes in acceptance and perception. A secondary aim was to compare these responses with attitudes among endoscopists at institutions where AI-assisted colonoscopy had not been used routinely.
Ethics statement
The study was reviewed by the SingHealth Centralised Institutional Review Board and was granted exemption from full review (No. 2022/2695). Ethical approval was not required because no patients were involved and no patient data were used. The study was reported according to the CHERRIES (Checklist for Reporting Results of Internet E-Surveys) statement [15].
Study design and setting
The survey was conducted 3 years after implementation of routine AI-assisted colonoscopy at SKH. Endoscopists at 7 tertiary healthcare institutions in Singapore, including SKH, were invited to participate. By that time, more clinicians across Singapore had had the opportunity to use AI systems, although not routinely. The survey was not limited to GI Genius users; it also included clinicians who had used other available computer-aided detection (CADe) and computer-aided characterization (CADx) systems. Both consultant-grade endoscopists and residents were eligible. Endoscopists were stratified by level of experience, including resident versus consultant status and total number of colonoscopies performed.
Survey
An online survey consisting of 44 English-language questions was developed to evaluate endoscopists’ knowledge and perceptions of AI, behaviors related to AI use in colonoscopy, and baseline demographic and professional characteristics. The questionnaire was based on an expanded version of the survey by Tham et al. [14], which was adapted from Mehta et al. [16]. The survey was administered through Google Forms (Google LLC).
The questionnaire was structured into 4 thematic domains: (1) knowledge of AI; (2) perceptions of what AI can accomplish in endoscopy; (3) perceptions of the personal and social impact of AI; and (4) behaviors related to AI-assisted tools in colonoscopy. A separate baseline section collected demographic and professional information, including the total number of colonoscopies performed to date, years of medical practice, current rank (e.g., consultant or resident), and primary workplace setting (e.g., public hospital, private practice, or academic institution). Question formats included 5-point Likert-scale items, including responses ranging from “strongly agree” to “strongly disagree” and from “extremely likely” to “extremely unlikely,” as well as multiple-choice and select-all-that-apply items.
To ensure respondent anonymity, all identifying information, including names, email addresses, IP addresses, and metadata automatically collected by the platform, was removed before data analysis. Only deidentified responses were included in the final dataset. Because responses were fully anonymized, participants could not be linked across survey rounds; therefore, the follow-up survey represented an independent, nonpaired cohort.
Statistical analysis
Descriptive statistics, including counts, medians, and interquartile ranges (IQRs), were calculated to summarize the data. For group comparisons, survey responses were categorized by combining “agree,” “strongly agree,” “likely,” and “extremely likely” responses to represent agreement. Differences in the proportion of respondents expressing agreement were assessed for each question using Fisher exact test. This approach was applied to comparisons between the initial and follow-up SKH surveys and to comparisons between SKH and other hospitals within the follow-up survey. Fisher exact test was selected because some groups had relatively small sample sizes. Statistical significance was defined as P<0.05. Analyses were not adjusted for AI system type because GI Genius was the predominant platform used among respondents. In an additional exploratory analysis, Likert-scale responses were analyzed as continuous variables by assigning numerical values from 1 (“strongly disagree” or “extremely unlikely”) to 5 (“strongly agree” or “extremely likely”). For each thematic domain, mean scores and standard deviations were calculated from the distribution of responses. Differences in mean domain scores between endoscopists from SKH and those from other hospitals were assessed using the independent 2-sample t-test. In addition, mean domain scores from SKH respondents in the present survey were compared with those from the previous survey to explore potential changes over time. Differences between groups were assessed using the independent two-sample t-test. All data analyses were performed using R ver. 4.2.1 (R Foundation for Statistical Computing).
From February to March 2024, a total of 57 endoscopists were included: 46 consultants (80.7%) and 11 residents (19.3%). The tabulated responses are shown in Tables 1–5 and Supplementary Tables 1–4. Overall, 35 endoscopists (61.4%) had performed more than 1,000 colonoscopies. All resident endoscopists had performed 1,000 or fewer colonoscopies. The median duration of medical practice was 14 years (IQR, 10–19 years). The most frequently reported type of AI system was CADe. Among specific systems, GI Genius was the most widely used, followed by CAD EYE (Fujifilm). In addition, 9 respondents (15.8%) reported having used more than 1 system.
Knowledge
Most responding clinicians were familiar with the term “artificial intelligence” (96.5%), and a large majority understood the concept of machine learning (87.7%). However, only 28 (49.1%) reported understanding neural networks, and 29 (50.9%) reported understanding deep learning (Table 1).
The mean knowledge domain score was 3.56±0.97 among endoscopists at SKH and 3.76±0.93 among endoscopists from other institutions (P=0.09) (Supplementary Table 1). In the previous survey, the mean knowledge score at SKH was 3.75±0.89 (P=0.17 for comparison with SKH in the present survey). Fig. 1 shows the mean Likert score for each survey question among endoscopists at SKH and those at other institutions in the present survey.
Perceptions
A large proportion of respondents believed that AI would raise new social challenges (98.2%) and new challenges related to health equity (91.2%) (Table 3). When asked whether current medical education adequately prepares clinicians to work alongside AI tools, 36 (63.2%) disagreed or strongly disagreed. In addition, 45 (78.9%) agreed that AI competency training should begin in medical school.
Agreement that AI training should be mandatory for all medical trainees was lower at SKH than at hospitals not routinely using AI (61.9% [13 of 21] vs. 88.9% [32 of 36], P=0.02) (Supplementary Table 3). In the follow-up survey, 19 of 21 SKH respondents (90.5%) agreed that AI would raise new challenges related to health equity, and 20 of 21 (95.2%) agreed that it would raise new social challenges; the corresponding proportions in the previous survey were 81.3% (13 of 16) and 87.5% (14 of 16), respectively. These differences were not statistically significant (P=0.63 and P=0.57, respectively). Agreement that the Singaporean healthcare system is well prepared to address AI-related challenges was 18.8% (3 of 16) in the first survey and 28.6% (6 of 21) in the follow-up survey (P=0.70). Among the responding endoscopists, 49 (86.0%) agreed that AI can be useful for evaluating the quality of a performed colonoscopy (Table 2). The mean score for perceptions of what AI can accomplish was 3.71±1.25 at SKH and 3.85±1.18 at other institutions (P=0.09) (Supplementary Table 2). In the previous survey, the mean score at SKH was 3.53±1.39 (P=0.08 for comparison with SKH in the present survey). The mean score for perceived personal and social impact of AI (questions 24–29) was 3.60±1.18 at SKH and 3.67±1.06 at other institutions (P=0.58). The mean score in the previous survey was 3.39±1.26 (P=0.21 for comparison with SKH in the present survey). The mean score for questions 30–34 was 3.36±1.17 at SKH and 3.52±1.10 among endoscopists at other institutions (P=0.25). The corresponding mean score in the previous survey was 3.22±1.12 (P=0.43 for comparison with the present survey).
Behaviors
Of the respondents, 39 (68.4%) reported that they enjoyed performing AI-assisted colonoscopies. In addition, 39 (68.4%) reported that they enjoyed performing AI-assisted colonoscopies. When respondents were stratified by number of colonoscopies performed, 74.3% of those who had performed more than 1,000 colonoscopies and 59.1% of those who had performed 1,000 or fewer reported enjoying the AI system. Most respondents (91.2%) agreed or strongly agreed that AI should be used as an adjunct to colonoscopy, including 97.1% of those who had performed more than 1,000 colonoscopies and 81.8% of those who had performed 1,000 or fewer (Table 4, Supplementary Table 4).
Among SKH endoscopists, agreement that AI should be used as an adjunct to colonoscopy was higher in the current survey than in the previous survey (95.2% [20 of 21] vs. 68.8% [11 of 16]), although the difference was not statistically significant (P=0.07). Agreement did not differ significantly between SKH and other hospitals in the current survey; 32 of 36 respondents at other hospitals (88.9%) agreed with this statement (P=0.64). When asked whether they enjoyed using AI assistance, 18 of 21 SKH respondents (85.7%) in the current survey agreed, compared with 11 of 16 (68.8%) in the previous survey (P=0.25). Among endoscopists at other hospitals, 21 of 36 (58.3%) reported enjoying AI assistance during colonoscopy, which was significantly lower than the proportion at SKH (P=0.04). In the current survey, 19 of 21 SKH endoscopists (90.5%) agreed that AI improves the quality of colonoscopy, compared with 12 of 16 (75.0%) in the previous survey (P=0.37). The corresponding proportion among endoscopists at other hospitals was 75.0% (27 of 36), with no significant difference compared with SKH (P=0.19). The mean score for this section was 4.17±0.58 at SKH and 3.88±0.81 at other institutions (P<0.01). The mean score in the previous SKH survey was 3.85±1.01 (P=0.05).
In this study, attitudes toward AI-assisted colonoscopy were evaluated over time among endoscopists with increasing experience using an AI-assisted system. The study also examined whether institutional differences, particularly the presence or absence of routine AI-assisted colonoscopy, were associated with endoscopists’ attitudes toward AI.
The gap between technical understanding and trust in AI
Clinicians appeared to be generally familiar with AI and reported high levels of acceptance of and trust in AI-assisted colonoscopy [17]. A substantial proportion of respondents indicated that current medical education does not adequately prepare clinicians to work with AI tools, and many believed that AI competency training should begin in medical school. These findings are consistent with respondents’ limited understanding of neural networks and deep learning and suggest a need for more comprehensive education on AI technologies in medical training.
Experience-based differences in behavior
Compared with the previous survey, a larger proportion of SKH respondents in the current survey believed that AI should be used as an adjunct to colonoscopy (95.2% vs. 68.8%) and reported enjoying AI assistance (85.7% vs. 68.8%). Although these differences may be clinically meaningful, they were not statistically significant, possibly because the sample size remained limited despite being larger than that of the previous survey. A borderline difference was observed when mean behavior scores from SKH endoscopists in the current survey were compared with those from the previous survey. The direction of these findings is consistent with recommendations from the Asia-Pacific Consensus Group, which emphasize the value of AI-assisted colonoscopy for improving polyp detection and characterization and for reducing advanced adenoma and colorectal cancer risk, while also noting the need for further validation as AI technologies evolve [18].
Endoscopists who had performed a larger number of colonoscopies were more likely to report enjoying the AI system and to believe that AI should be included as an adjunct to colonoscopy. This finding is consistent with the previous survey. Another survey found that endoscopists with more years of experience appeared more willing to accept risk and trust AI than those with less experience, possibly because they viewed AI as a supportive tool rather than a requirement [17].
Perceptions of the impact of AI
Only a minority of respondents agreed that the Singaporean healthcare system is well prepared to address AI-related challenges: 18.8% in the first survey and 28.6% in the follow-up survey, with no statistically significant difference between surveys.
Belief that AI would raise new social and health equity challenges appeared to increase between the initial and follow-up surveys at SKH, although the differences were not statistically significant. Regional variation in cost-effectiveness should be considered when implementing this technology [18]. AI technology remains relatively expensive, which may create disparities if less economically developed countries have fewer opportunities than more developed countries to invest in and benefit from these tools. On the basis of 2 simulation studies, the World Endoscopy Organization stated that CADe is likely to increase healthcare costs in the short term because increased polyp removal may lead to additional surveillance colonoscopies; however, this increase could be offset in the long term by lower cancer-treatment costs resulting from improved cancer prevention [19–21]. CADx may further reduce healthcare costs by decreasing unnecessary removal and pathological examination of diminutive polyps (≤5 mm), provided that diagnostic performance is sufficient [19, 22]. These assumptions are derived from modeling studies, which should be considered when applying the findings to real-world practice. More reliable estimates will depend on long-term clinical outcome data rather than simulation-based modeling alone. Geographic variation may also not be fully captured in these results [19]. Endoscopists have also expressed concerns about AI-related costs [23]. To address implementation challenges, healthcare systems and authorities should consider both the clinical benefits and cost-effectiveness of AI tools [19].
Institutional variation
Institutional variation was observed in behaviors related to AI assistance during colonoscopy. When SKH was compared with other institutions where endoscopists had less experience with AI-assisted colonoscopy, a larger proportion of SKH endoscopists agreed that AI should be used as an adjunct to colonoscopy and that it improves procedural quality, although these differences were not statistically significant. When respondents were asked whether they enjoyed AI-assisted colonoscopy, only 58.3% of endoscopists at other hospitals agreed, compared with 85.7% at SKH; this difference was statistically significant. Overall, mean domain scores were broadly similar between endoscopists at SKH and those from other institutions. However, the significant difference in behavior scores further suggests a possible positive association between greater AI exposure and more favorable attitudes.
Strengths and limitations
A key strength of this study is its follow-up design, which allowed group-level assessment of how endoscopists’ attitudes toward AI changed over time with increasing hands-on experience. This approach provides insight into the process of AI acceptance in clinical practice.
This study has several limitations. Although the sample size increased from 16 to 57 participants compared with the original survey, it remained relatively small and may have limited the statistical power to detect significant differences. As a result, no significant differences were identified between SKH responses across the 2 surveys. The use of independent, nonpaired cohorts limited the ability to assess individual-level changes over time. The survey also lacked a specific AI-exposure-based analysis because respondents were not stratified by actual exposure to AI assistance. This limitation makes it difficult to distinguish the effects of AI familiarity from those of general professional seniority or institutional culture. In addition, all conclusions were based on subjective self-reported attitudes without linkage to objective performance metrics such as ADR. Future studies correlating attitudes with objective performance measures could provide additional insight.
The domain-level comparisons should be interpreted cautiously. Although the overall questionnaire structure was maintained, several additional questions were included in the follow-up survey. As a result, the domain-level aggregates were not based on completely identical item sets, limiting direct one-to-one comparisons between surveys. Therefore, the domain-level analysis should be considered exploratory and primarily intended to provide a general overview of trends rather than definitive evidence of changes over time.
Another limitation is that the survey response rate was not tracked, which may have introduced selection bias, particularly toward clinicians with favorable views of AI, and may have affected the representativeness of the findings. Without response-rate data, it is difficult to determine whether the respondents were representative of the broader endoscopist population, raising the possibility of nonresponse bias. A mixed-methods study incorporating both qualitative interviews and quantitative survey data would provide a more comprehensive understanding of endoscopists’ perspectives. Although the survey captured broad trends and patterns, interviews could provide deeper insight into the reasoning, experiences, and contextual factors that influence attitudes and behaviors toward AI in healthcare and colonoscopy. Another limitation is that the survey was conducted in an affluent Asian setting, which may not reflect perspectives across the broader Asian region or Southeast Asia. Future studies should include a more diverse regional sample to better address issues of health equity and equality. Future surveys should also assess cost-effectiveness because it is an important factor in the adoption and implementation of AI in colonoscopy.
Conclusions
AI was generally perceived as improving the quality of colonoscopy, although some respondents did not enjoy AI-assisted colonoscopy. The findings indicate that greater colonoscopy experience and more extensive exposure to AI-assisted colonoscopy were associated with more positive attitudes toward using AI as an adjunct to colonoscopy and greater acceptance of its potential to improve procedural quality.

Conflict of interest

Frederick H. Koh is a Key-Opinion-Leader for Medtronic for AI in Endoscopy, Asia-Pacific Region. He is also an editorial board member of this journal, but was not involved in the peer reviewer selection, evaluation, or decision process of this article. No other potential conflict of interest relevant to this article was reported.

Funding

This study was supported by the Danish Cancer Society, which funded part of Ronja M. B. Lagström’s research stay in Singapore during her PhD.

Author contributions

Conceptualization: all authors; Data curation: RMBL; Formal analysis: RMBL; Funding acquisition: RMBL; Investigation: RMBL; Methodology: FHK, RMBL, IG; Project administration: FHK; Visualization: RMBL; Writing–original draft: RMBL; Writing–review & editing: all authors. All authors read and approved the final manuscript.

Supplementary Material 1.

The full list of the members of SKH Endoscopy Centre and Artificial Intelligence in Gastrointestinal Endoscopy Singapore (AIGES).
ac-2026-00178-0025-Supplementary-Material-1.pdf

Supplementary Table 1.

Survey responses on knowledge of AI (questions 1–5), stratified based on their number of performed colonoscopies and institution
ac-2026-00178-0025-Supplementary-Table-1.pdf

Supplementary Table 2.

Survey responses on perception of what AI can accomplish (questions 6–23), stratified based on their number of performed colonoscopies and institution
ac-2026-00178-0025-Supplementary-Table-2.pdf

Supplementary Table 3.

Survey responses on perceptions of the personal and social impact of AI (questions 24–35), stratified based on their number of performed colonoscopies and institution
ac-2026-00178-0025-Supplementary-Table-3.pdf

Supplementary Table 4.

Survey responses on behaviors pertaining to use of AI assistance in colonoscopy (questions 36–40), stratified based on their number of performed colonoscopies and institution
ac-2026-00178-0025-Supplementary-Table-4.pdf
Supplementary materials are available from https://doi.org/10.3393/ac.2026.00178.0025.
Fig. 1.
Radar plot showing the mean Likert score for each survey question (Q1–Q34 and Q38–Q40) among endoscopists at Sengkang General Hospital (SKH) and other institutions.
ac-2026-00178-0025f1.jpg
ac-2026-00178-0025f2.jpg
Table 1.
Knowledge of artificial intelligence (questions 1–5) (n=57)
No. Question No. of responses (%)
Strongly agree Agree Neither Disagree Strongly disagree
1. I understand what the term “artificial intelligence” means 13 (22.8) 42 (73.7) 2 (3.5) 0 (0) 0 (0)
2. I understand what the term “machine learning” means 10 (17.5) 40 (70.2) 4 (7.0) 3 (5.3) 0 (0)
3. I understand what the term “neural network” means 8 (14.0) 20 (35.1) 13 (22.8) 15 (26.3) 1 (1.8)
4. I understand what the term “deep learning” means 5 (8.8) 24 (42.1) 13 (22.8) 14 (24.6) 1 (1.8)
5. I know what an algorithm is in the context of computer science 6 (10.5) 36 (63.2) 4 (7.0) 8 (14.0) 3 (5.3)
Table 2.
Perceptions of what artificial intelligence can accomplish (questions 6–23) (n=57)
No. Question No. of responses (%)
Extremely likely Likely Uncertain Unlikely Extremely unlikely
6. Provide patients with preventive health recommendations (e.g., exercise, diet, wellness) 28 (49.1) 28 (49.1) 0 (0) 1 (1.8) 0 (0)
7. Analyze patient information to reach diagnoses 22 (38.6) 28 (49.1) 3 (5.3) 4 (7.0) 0 (0)
8. Analyze patient information to establish prognoses 28 (49.1) 24 (42.1) 3 (5.3) 2 (3.5) 0 (0)
9. Read and interpret diagnostic imaging 34 (59.6) 23 (40.4) 0 (0) 0 (0) 0 (0)
10. Monitor patient compliance with prescribed medications, exercise and dietary recommendations 24 (42.1) 19 (33.3) 3 (5.3) 10 (17.5) 1 (1.8)
11. Evaluate when to refer patients to other health professionals 8 (14.0) 33 (57.9) 4 (7.0) 9 (15.8) 3 (5.3)
12. Formulate personalized treatment plans for patients 7 (12.3) 29 (50.9) 4 (7.0) 14 (24.6) 3 (5.3)
13. Formulate personalized medication prescriptions for patients 9 (15.8) 31 (54.4) 4 (7.0) 11 (19.3) 2 (3.5)
14. Provide empathetic care to patients 3 (5.3) 8 (14.0) 1 (1.8) 26 (45.6) 19 (33.3)
15. Provide psychiatric/personal counseling 2 (3.5) 11 (19.3) 4 (7.0) 26 (45.6) 14 (24.6)
16. Perform surgery (e.g., robotic surgery) 6 (10.5) 22 (38.6) 1 (1.8) 19 (33.3) 9 (15.8)
17. Image-guided radiotherapy 14 (24.6) 32 (56.1) 1 (1.8) 9 (15.8) 1 (1.8)
18. Assess competency/evaluate the quality of a performed colonoscopy 23 (40.4) 26 (45.6) 1 (1.8) 6 (10.5) 1 (1.8)
19. Provide documentation (e.g., update medical records) about patients 22 (38.6) 32 (56.1) 1 (1.8) 2 (3.5) 0 (0)
20. Assist hospitals in capacity planning and human resource management 36 (63.2) 16 (28.1) 0 (0) 4 (7.0) 1 (1.8)
21. Provide recommendations for quality improvement in practices/hospitals 23 (40.4) 25 (43.9) 0 (0) 6 (10.5) 3 (5.3)
22. Conduct population health surveillance and outbreak prevention 27 (47.4) 26 (45.6) 1 (1.8) 3 (5.3) 0 (0)
23. Select the best population health interventions 12 (21.1) 30 (52.6) 3 (5.3) 10 (17.5) 2 (3.5)
Table 3.
Perceptions of the personal and social impact of AI (questions 24–35) (n=57)
No. Question No. of responses (%)
Strongly agree Agree Neither Disagree Strongly disagree
24. AI will reduce the number of jobs available to me 0 (0) 17 (29.8) 11 (19.3) 28 (49.1) 1 (1.8)
25. AI will reduce the number of jobs in certain medical specialties more than others 6 (10.5) 41 (71.9) 4 (7.0) 6 (10.5) 0 (0)
26. AI will/already did impact my choice of specialty selection 1 (1.8) 6 (10.5) 11 (19.3) 32 (56.1) 7 (12.3)
27. AI in medicine will raise new ethical challenges 24 (42.1) 31 (54.4) 1 (1.8) 0 (0) 1 (1.8)
28. AI in medicine will raise new social challenges 22 (38.6) 34 (59.6) 1 (1.8) 0 (0) 0 (0)
29. AI in medicine will raise new challenges around health equity 19 (33.3) 33 (57.9) 2 (3.5) 3 (5.3) 0 (0)
30. The Singaporean healthcare system is currently well prepared to deal with challenges having to do with AI 3 (5.3) 15 (26.3) 22 (38.6) 14 (24.6) 3 (5.3)
31. Knowledge of how the AI algorithm is trained is essential in the implementation of the system 16 (28.1) 29 (50.9) 6 (10.5) 4 (7.0) 2 (3.5)
32. My medical education is adequately preparing me for working alongside AI tools 0 (0) 9 (15.8) 12 (21.1) 26 (45.6) 10 (17.5)
33. Medical training should include training on AI competencies (e.g. what is AI, how will it impact us, what are the challenges it raises) 16 (28.1) 31 (54.4) 8 (14.0) 2 (3.5) 0 (0)
34. Every medical trainee should be required to receive training in AI competencies 13 (22.8) 32 (56.1) 8 (14.0) 4 (7.0) 0 (0)
35.a Training in AI competencies should begin as a:
 Medical student 45 (78.9)
 Resident 4 (7.0)
 Practicing physician 4 (7.0)
 Not necessary 4 (7.0)

AI, artificial intelligence.

aFor question 35, instead of a 5-point Likert scale, response categories were medical student, resident, practicing physician, and not necessary.

Table 4.
Behaviors pertaining to use of AI assistance in colonoscopy (questions 36–40) (n=57)
No. Question No. of responses (%)
36.a Which system(s) have you been using (e.g., GI Genius, CAD EYE)?
 GI Genius (Medtronic) 38 (66.7)
 CAD EYE (Fujifilm) 22 (38.6)
 ENDO-AID (Olympus Corp) 4 (7.0)
 EndoAngel (EndoAngel) 2 (3.5)
 NA/missing 6 (10.5)
37. Which type of AI system have you been using the most in your practice?
 CADe 50 (87.7)
 CADx 7 (12.3)
 CAQ 0 (0)
38. Did you enjoy using the AI system during colonoscopy?
 Strongly agree 9 (15.8)
 Agree 30 (52.6)
 Neither 15 (26.3)
 Disagree 3 (5.3)
 Strongly disagree 0 (0)
39. AI should be included as an adjunct in colonoscopy. Do you agree?
 Strongly agree 16 (28.1)
 Agree 36 (63.2)
 Neither 3 (5.3)
 Disagree 2 (3.5)
 Strongly disagree 0 (0)
40. AI improves the quality of colonoscopy. Do you agree?
 Strongly agree 14 (24.6)
 Agree 32 (56.1)
 Neither 9 (15.8)
 Disagree 2 (3.5)
 Strongly disagree 0 (0)

AI, artificial intelligence; NA, not applicable; CADe, computer-aided detection; CADx, computer-aided characterization; CAQ, computer-aided quality assessment

aThe sum of responses may exceed the total number of respondents because multiple responses were permitted.

Table 5.
Baseline information on the endoscopists (questions 41–44) (n=57)
No. Question No. of responses (%)a
41. How many colonoscopies have you performed?
 ≤1,000 22 (38.6)
 >1,000 35 (61.4)
42. For how many years have you practiced medicine? (median [IQR]) 14 (10–19)
43. What is your rank?
 Consultant 46 (80.7)
 Resident 11 (19.3)
44. Where do you work?
 Sengkang General Hospital 21 (36.8)
 Changi General Hospital 17 (29.8)
 Other 19 (33.3)

IQR, interquartile range.

aUnless otherwise indicated.

  • 1. Milluzzo SM, Cesaro P, Grazioli LM, Olivari N, Spada C. Artificial intelligence in lower gastrointestinal endoscopy: the current status and future perspective. Clin Endosc 2021;54:329–39. ArticlePubMedPMCPDF
  • 2. Hassan C, Wallace MB, Sharma P, Maselli R, Craviotto V, Spadaccini M, et al. New artificial intelligence system: first validation study versus experienced endoscopists for colorectal polyp detection. Gut 2020;69:799–800. ArticlePubMed
  • 3. Repici A, Badalamenti M, Maselli R, Correale L, Radaelli F, Rondonotti E, et al. Efficacy of real-time computer-aided detection of colorectal neoplasia in a randomized trial. Gastroenterology 2020;159:512–20.e7. ArticlePubMed
  • 4. Repici A, Spadaccini M, Antonelli G, Correale L, Maselli R, Galtieri PA, et al. Artificial intelligence and colonoscopy experience: lessons from two randomised trials. Gut 2022;71:757–65. ArticlePubMed
  • 5. Gong D, Wu L, Zhang J, Mu G, Shen L, Liu J, et al. Detection of colorectal adenomas with a real-time computer-aided system (ENDOANGEL): a randomised controlled study. Lancet Gastroenterol Hepatol 2020;5:352–61. ArticlePubMed
  • 6. Wang P, Liu X, Berzin TM, Glissen Brown JR, Liu P, Zhou C, et al. Effect of a deep-learning computer-aided detection system on adenoma detection during colonoscopy (CADe-DB trial): a double-blind randomised study. Lancet Gastroenterol Hepatol 2020;5:343–51. ArticlePubMed
  • 7. Liu WN, Zhang YY, Bian XQ, Wang LJ, Yang Q, Zhang XD, et al. Study on detection rate of polyps and adenomas in artificial-intelligence-aided colonoscopy. Saudi J Gastroenterol 2020;26:13–9. ArticlePubMedPMC
  • 8. Su JR, Li Z, Shao XJ, Ji CR, Ji R, Zhou RC, et al. Impact of a real-time automatic quality control system on colorectal polyp and adenoma detection: a prospective randomized controlled study (with videos). Gastrointest Endosc 2020;91:415–24.e4. ArticlePubMed
  • 9. Wang P, Berzin TM, Glissen Brown JR, Bharadwaj S, Becq A, Xiao X, et al. Real-time automatic detection system increases colonoscopic polyp and adenoma detection rates: a prospective randomised controlled study. Gut 2019;68:1813–9. ArticlePubMedPMC
  • 10. Liu P, Wang P, Glissen Brown JR, Berzin TM, Zhou G, Liu W, et al. The single-monitor trial: an embedded CADe system increased adenoma detection during colonoscopy: a prospective randomized study. Ther Adv Gastroenterol 2020;13:1756284820979165.ArticlePubMedPMCPDF
  • 11. Lagström RM, Bräuner KB, Bielik J, Rosen AW, Crone JG, Gögenur I, et al. Improvement in adenoma detection rate by artificial intelligence-assisted colonoscopy: multicenter quasi-randomized controlled trial. Endosc Int Open 2025;13:a25215169. ArticlePMC
  • 12. Koh FH, Ladlad J; SKH Endoscopy Centre, Teo EK, Lin CL, Foo FJ. Real-time artificial intelligence (AI)-aided endoscopy improves adenoma detection rates even in experienced endoscopists: a cohort study in Singapore. Surg Endosc 2023;37:165–71. ArticlePubMed
  • 13. Chin SE, Wan FT, Ladlad J, Chue KM; SKH Endoscopy Centre, Teo EK, et al. One-year review of real-time artificial intelligence (AI)-aided endoscopy performance. Surg Endosc 2023;37:6402–7. ArticlePubMedPDF
  • 14. Tham S, Koh FH; SKH Endoscopy Centre, Teo EK, Lin CL, Foo FJ. Knowledge, perceptions and behaviours of endoscopists towards the use of artificial intelligence-aided colonoscopy. Surg Endosc 2023;37:7395–400. ArticlePubMed
  • 15. Eysenbach G. Improving the quality of web surveys: the Checklist for Reporting Results of Internet E-Surveys (CHERRIES). J Med Internet Res 2004;6:e34. ArticlePubMedPMC
  • 16. Mehta N, Harish V, Bilimoria K, Morgado F, Ginsburg S, Law M, et al. Knowledge and attitudes on artificial intelligence in healthcare: a provincial survey study of medical students. MedEdPublish 2021;10:75.Article
  • 17. Goh WW, Chia KY, Cheung MF, Kee KM, Lwin MO, Schulz PJ, et al. Risk perception, acceptance, and trust of using AI in gastroenterology practice in the Asia-Pacific region: web-based survey study. JMIR AI 2024;3:e50525. ArticlePubMedPMC
  • 18. Koh FH, Li JW, Wong SH, Lee J, John S, Chong VH, et al. Asia-Pacific consensus on the use of artificial intelligence in colorectal cancer screening and surveillance. IGIE 2025;4:390–9. ArticlePubMedPMC
  • 19. Mori Y, East JE, Hassan C, Halvorsen N, Berzin TM, Byrne M, et al. Benefits and challenges in implementation of artificial intelligence in colonoscopy: World Endoscopy Organization position statement. Dig Endosc 2023;35:422–9. ArticlePubMedPMCPDF
  • 20. Areia M, Mori Y, Correale L, Repici A, Bretthauer M, Sharma P, et al. Cost-effectiveness of artificial intelligence for screening colonoscopy: a modelling study. Lancet Digit Health 2022;4:e436–44. ArticlePubMed
  • 21. Mori Y, Wang P, Løberg M, Misawa M, Repici A, Spadaccini M, et al. Impact of artificial intelligence on colonoscopy surveillance after polyp removal: a pooled analysis of randomized trials. Clin Gastroenterol Hepatol 2023;21:949–59. e2. ArticlePubMed
  • 22. Rex DK, Kahi C, O'Brien M, Levin TR, Pohl H, Rastogi A, et al. The American Society for Gastrointestinal Endoscopy PIVI (preservation and incorporation of valuable endoscopic innovations) on real-time endoscopic assessment of the histology of diminutive colorectal polyps. Gastrointest Endosc 2011;73:419–22. ArticlePubMed
  • 23. Wadhwa V, Alagappan M, Gonzalez A, Gupta K, Brown JR, Cohen J, et al. Physician sentiment toward artificial intelligence (AI) in colonoscopic practice: a survey of US gastroenterologists. Endosc Int Open 2020;8:E1379–84. ArticlePubMedPMC

Figure & Data

References

    Citations

    Citations to this article as recorded by  

      • Cite this Article
        Cite this Article
        export Copy Download
        Close
        Download Citation
        Download a citation file in RIS format that can be imported by all major citation management software, including EndNote, ProCite, RefWorks, and Reference Manager.

        Format:
        • RIS — For EndNote, ProCite, RefWorks, and most other reference management software
        • BibTeX — For JabRef, BibDesk, and other BibTeX-specific software
        Include:
        • Citation for the content below
        Endoscopists’ attitudes toward and experiences with artificial intelligence–assisted colonoscopy: a 3-year follow-up survey after the implementation of GI Genius
        Ann Coloproctol. 2026;42(4):436-445.   Published online August 25, 2026
        Close
      • XML DownloadXML Download
      Figure
      • 0
      • 1
      Endoscopists’ attitudes toward and experiences with artificial intelligence–assisted colonoscopy: a 3-year follow-up survey after the implementation of GI Genius
      Image Image
      Fig. 1. Radar plot showing the mean Likert score for each survey question (Q1–Q34 and Q38–Q40) among endoscopists at Sengkang General Hospital (SKH) and other institutions.
      Graphical abstract
      Endoscopists’ attitudes toward and experiences with artificial intelligence–assisted colonoscopy: a 3-year follow-up survey after the implementation of GI Genius
      No. Question No. of responses (%)
      Strongly agree Agree Neither Disagree Strongly disagree
      1. I understand what the term “artificial intelligence” means 13 (22.8) 42 (73.7) 2 (3.5) 0 (0) 0 (0)
      2. I understand what the term “machine learning” means 10 (17.5) 40 (70.2) 4 (7.0) 3 (5.3) 0 (0)
      3. I understand what the term “neural network” means 8 (14.0) 20 (35.1) 13 (22.8) 15 (26.3) 1 (1.8)
      4. I understand what the term “deep learning” means 5 (8.8) 24 (42.1) 13 (22.8) 14 (24.6) 1 (1.8)
      5. I know what an algorithm is in the context of computer science 6 (10.5) 36 (63.2) 4 (7.0) 8 (14.0) 3 (5.3)
      No. Question No. of responses (%)
      Extremely likely Likely Uncertain Unlikely Extremely unlikely
      6. Provide patients with preventive health recommendations (e.g., exercise, diet, wellness) 28 (49.1) 28 (49.1) 0 (0) 1 (1.8) 0 (0)
      7. Analyze patient information to reach diagnoses 22 (38.6) 28 (49.1) 3 (5.3) 4 (7.0) 0 (0)
      8. Analyze patient information to establish prognoses 28 (49.1) 24 (42.1) 3 (5.3) 2 (3.5) 0 (0)
      9. Read and interpret diagnostic imaging 34 (59.6) 23 (40.4) 0 (0) 0 (0) 0 (0)
      10. Monitor patient compliance with prescribed medications, exercise and dietary recommendations 24 (42.1) 19 (33.3) 3 (5.3) 10 (17.5) 1 (1.8)
      11. Evaluate when to refer patients to other health professionals 8 (14.0) 33 (57.9) 4 (7.0) 9 (15.8) 3 (5.3)
      12. Formulate personalized treatment plans for patients 7 (12.3) 29 (50.9) 4 (7.0) 14 (24.6) 3 (5.3)
      13. Formulate personalized medication prescriptions for patients 9 (15.8) 31 (54.4) 4 (7.0) 11 (19.3) 2 (3.5)
      14. Provide empathetic care to patients 3 (5.3) 8 (14.0) 1 (1.8) 26 (45.6) 19 (33.3)
      15. Provide psychiatric/personal counseling 2 (3.5) 11 (19.3) 4 (7.0) 26 (45.6) 14 (24.6)
      16. Perform surgery (e.g., robotic surgery) 6 (10.5) 22 (38.6) 1 (1.8) 19 (33.3) 9 (15.8)
      17. Image-guided radiotherapy 14 (24.6) 32 (56.1) 1 (1.8) 9 (15.8) 1 (1.8)
      18. Assess competency/evaluate the quality of a performed colonoscopy 23 (40.4) 26 (45.6) 1 (1.8) 6 (10.5) 1 (1.8)
      19. Provide documentation (e.g., update medical records) about patients 22 (38.6) 32 (56.1) 1 (1.8) 2 (3.5) 0 (0)
      20. Assist hospitals in capacity planning and human resource management 36 (63.2) 16 (28.1) 0 (0) 4 (7.0) 1 (1.8)
      21. Provide recommendations for quality improvement in practices/hospitals 23 (40.4) 25 (43.9) 0 (0) 6 (10.5) 3 (5.3)
      22. Conduct population health surveillance and outbreak prevention 27 (47.4) 26 (45.6) 1 (1.8) 3 (5.3) 0 (0)
      23. Select the best population health interventions 12 (21.1) 30 (52.6) 3 (5.3) 10 (17.5) 2 (3.5)
      No. Question No. of responses (%)
      Strongly agree Agree Neither Disagree Strongly disagree
      24. AI will reduce the number of jobs available to me 0 (0) 17 (29.8) 11 (19.3) 28 (49.1) 1 (1.8)
      25. AI will reduce the number of jobs in certain medical specialties more than others 6 (10.5) 41 (71.9) 4 (7.0) 6 (10.5) 0 (0)
      26. AI will/already did impact my choice of specialty selection 1 (1.8) 6 (10.5) 11 (19.3) 32 (56.1) 7 (12.3)
      27. AI in medicine will raise new ethical challenges 24 (42.1) 31 (54.4) 1 (1.8) 0 (0) 1 (1.8)
      28. AI in medicine will raise new social challenges 22 (38.6) 34 (59.6) 1 (1.8) 0 (0) 0 (0)
      29. AI in medicine will raise new challenges around health equity 19 (33.3) 33 (57.9) 2 (3.5) 3 (5.3) 0 (0)
      30. The Singaporean healthcare system is currently well prepared to deal with challenges having to do with AI 3 (5.3) 15 (26.3) 22 (38.6) 14 (24.6) 3 (5.3)
      31. Knowledge of how the AI algorithm is trained is essential in the implementation of the system 16 (28.1) 29 (50.9) 6 (10.5) 4 (7.0) 2 (3.5)
      32. My medical education is adequately preparing me for working alongside AI tools 0 (0) 9 (15.8) 12 (21.1) 26 (45.6) 10 (17.5)
      33. Medical training should include training on AI competencies (e.g. what is AI, how will it impact us, what are the challenges it raises) 16 (28.1) 31 (54.4) 8 (14.0) 2 (3.5) 0 (0)
      34. Every medical trainee should be required to receive training in AI competencies 13 (22.8) 32 (56.1) 8 (14.0) 4 (7.0) 0 (0)
      35.a Training in AI competencies should begin as a:
       Medical student 45 (78.9)
       Resident 4 (7.0)
       Practicing physician 4 (7.0)
       Not necessary 4 (7.0)
      No. Question No. of responses (%)
      36.a Which system(s) have you been using (e.g., GI Genius, CAD EYE)?
       GI Genius (Medtronic) 38 (66.7)
       CAD EYE (Fujifilm) 22 (38.6)
       ENDO-AID (Olympus Corp) 4 (7.0)
       EndoAngel (EndoAngel) 2 (3.5)
       NA/missing 6 (10.5)
      37. Which type of AI system have you been using the most in your practice?
       CADe 50 (87.7)
       CADx 7 (12.3)
       CAQ 0 (0)
      38. Did you enjoy using the AI system during colonoscopy?
       Strongly agree 9 (15.8)
       Agree 30 (52.6)
       Neither 15 (26.3)
       Disagree 3 (5.3)
       Strongly disagree 0 (0)
      39. AI should be included as an adjunct in colonoscopy. Do you agree?
       Strongly agree 16 (28.1)
       Agree 36 (63.2)
       Neither 3 (5.3)
       Disagree 2 (3.5)
       Strongly disagree 0 (0)
      40. AI improves the quality of colonoscopy. Do you agree?
       Strongly agree 14 (24.6)
       Agree 32 (56.1)
       Neither 9 (15.8)
       Disagree 2 (3.5)
       Strongly disagree 0 (0)
      No. Question No. of responses (%)a
      41. How many colonoscopies have you performed?
       ≤1,000 22 (38.6)
       >1,000 35 (61.4)
      42. For how many years have you practiced medicine? (median [IQR]) 14 (10–19)
      43. What is your rank?
       Consultant 46 (80.7)
       Resident 11 (19.3)
      44. Where do you work?
       Sengkang General Hospital 21 (36.8)
       Changi General Hospital 17 (29.8)
       Other 19 (33.3)
      Table 1. Knowledge of artificial intelligence (questions 1–5) (n=57)

      Table 2. Perceptions of what artificial intelligence can accomplish (questions 6–23) (n=57)

      Table 3. Perceptions of the personal and social impact of AI (questions 24–35) (n=57)

      AI, artificial intelligence.

      For question 35, instead of a 5-point Likert scale, response categories were medical student, resident, practicing physician, and not necessary.

      Table 4. Behaviors pertaining to use of AI assistance in colonoscopy (questions 36–40) (n=57)

      AI, artificial intelligence; NA, not applicable; CADe, computer-aided detection; CADx, computer-aided characterization; CAQ, computer-aided quality assessment

      The sum of responses may exceed the total number of respondents because multiple responses were permitted.

      Table 5. Baseline information on the endoscopists (questions 41–44) (n=57)

      IQR, interquartile range.

      Unless otherwise indicated.


      Ann Coloproctol : Annals of Coloproctology Twitter Facebook
      TOP