Human–machine interaction in optical polyp diagnosis: decision-making after CADx polyp diagnosis
Open accessAbstract
Background: Artificial intelligence (AI)-based computer-aided diagnosis (CADx) systems are increasingly used to classify colorectal polyps during colonoscopy, but their impact depends on how endoscopists interpret AI predictions. This study aimed to evaluate endoscopists’ decision-making after reviewing CADx outputs, using histopathology diagnosis as ground truth. Patients and methods: We performed a secondary analysis of a prospective study (NCT06822816) conducted at our centre between 2022 and 2025. All patients and polyps ≤10 mm with documented CADx output and endoscopist diagnosis were included. Endoscopists either accepted CADx classifications (neoplastic vs hyperplastic) or provided alternatives. Our primary outcome was the accuracy of accepted CADx diagnoses; secondary outcomes assessed reclassification patterns after CADx output rejection. Results: Our cohort included 754 polyps <10 mm from 367 patients. Accepted CADx diagnoses were correct for 89.1% of neoplastic and 68.7% of hyperplastic CADx predictions. Endoscopists accepted incorrect CADx classifications more frequently when these were neoplastic compared to hyperplastic (64.9% vs 26.5%). Upon rejecting correct CADx neoplastic predictions, endoscopists misclassified 23.4% as sessile serrated lesions (SSLs). Similarly, 29.5% of rejected correct CADx hyperplastic predictions were incorrectly reassigned as SSLs. Conclusions: Endoscopists exhibit a diagnostic bias toward neoplasia, particularly misclassifying hyperplastic polyps as SSLs. While this may enhance safety, it may risk unnecessary surveillance colonoscopies and increase costs.
AI snapshot
Written from the paper’s full text.
TL;DR
In this secondary analysis of a prospective CADx colonoscopy cohort, endoscopists accepted correct neoplastic CADx diagnoses more often than correct hyperplastic ones and showed a bias toward neoplasia, often reclassifying rejected hyperplastic predictions as SSLs. This conservative bias may be clinically safe but could increase unnecessary surveillance colonoscopies and costs.
Key findings
- Accepted CADx diagnoses were correct for 89.1% of neoplastic predictions (408/458; 95% CI, 85.8–91.7) and 68.7% of hyperplastic predictions (68/99; 95% CI, 58.5–77.4), P < .001.
- Endoscopists accepted incorrect CADx classifications more frequently when the incorrect classification was neoplastic (64.9% of 77 cases) than when it was hyperplastic (26.5% of 117 cases).
- Post-rejection reclassification showed a bias toward SSL: the abstract reports 23.4% of rejected correct neoplastic predictions and 29.5% of rejected correct hyperplastic predictions were misclassified as SSLs; the Results report 53.2% (25/47) of rejected neoplastic predictions and 41.6% (62/149) of rejected hyperplastic outputs were incorrectly reassigned as SSLs.
- Location subgroup: in the proximal colon, correctly accepted CADx neoplastic diagnoses were 91.1% vs 34.6% for hyperplastic (P < .001); in the distal colon, the proportions were not significantly different (84.8% vs 80.8%, P = .577).
Methods
- Secondary analysis of an IRB-approved prospective cohort (NCT06822816) conducted at one centre between June 2022 and January 2025, using the CADEYE characterization system (Fujifilm Co.) during colonoscopy.
- Included 753 polyps ≤10 mm from 367 patients with documented CADx output, endoscopist histology prediction, and adenoma, hyperplastic, or SSL pathology; excluded polyps without CADx diagnosis and polyps >10 mm. Colonoscopies were performed by 16 endoscopists (8 staff, N = 180; 8 trainees, N = 187).
- Primary outcome was proportion of correctly accepted CADx diagnoses stratified by neoplastic vs hyperplastic; secondary outcomes included correct rejection of inaccurate CADx outputs and reclassification patterns. Proportions reported with 95% CIs; comparisons used chi-squared or Fisher exact tests; analyses in R version 4.4.2.
Limitations
- The paper states a key limitation is the mismatch between binary CADx outputs (neoplastic vs hyperplastic) and endoscopists’ three-category scheme (adenoma, hyperplastic, SSL), since current CADx systems cannot classify SSLs; many apparent reclassification errors may reflect semantic misalignment rather than human misjudgment.
- The paper states inclusion was limited to polyps <10 mm, and diagnostic accuracy for both endoscopists and CADx declines for smaller polyps, which may explain variability in final diagnostic accuracy.
- The paper states lack of pathologist data, including no Kappa among pathologists interpreting hyperplastic polyps and SSLs, could confound the histopathology gold standard. Evident from the text: this is a single-centre secondary analysis, and the abstract reports 754 polyps while the Results and Table 1 report 753; the abstract and Results also give inconsistent post-rejection SSL reclassification percentages.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Claims the paper makes
1 of 5 cited sources checked · 0 supported · 0 flagged · accuracy not yet measured
- Interpretationcitation
The conservative approach of classifying benign lesions as neoplastic may inadvertently lead to unnecessary early surveillance colonoscopies, thus offsetting initial pathology cost-savings.
“However, this conservative approach may inadvertently lead to unnecessary early surveillance colonoscopies, thus offsetting initial pathology cost-savings.12,13”
Quote in the paperSubstantial claim1 of 2 sources checked- cites [12] Force, 2025, Consensus statements on the current landscape of artificial intelligence applications in endoscopy, addressing roadblocks, and advancing artificial intelligence in gastroenterology, Gastrointest Endosc source
The discussion of AI in daily endoscopy practice says algorithms can recommend appropriate surveillance intervals after polyp resection, but it does not address benign lesions being classified as neoplastic or the resulting costs. Searches for early surveillance, false-positive classification, and pathology cost-savings found no passage in our copy supporting the claimed trade-off. (LLM, conf=0.73)
Experimental — accuracy not yet measured - cites [13] Halvorsen, 2025, Leaving colorectal polyps in situ with endocytoscopy assisted by computer-aided diagnosis: a cost-effectiveness study, Endoscopy source
- Findingown data
The proportion of correctly accepted CADx neoplastic diagnoses was greater than the proportion of correctly accepted CADx hyperplastic diagnoses (89.1% vs 68.7%, P < .001).
“The proportion of correctly accepted CADx neoplastic diagnoses (89.1%, 408/458; 95% CI, 85.8-91.7, Figure 2) was greater than the proportion of correctly accepted CADx hyperplastic diagnoses (68.7%, 68/99; 95% CI, 58.5-77.4, Figure 3), P < .001.”
Quote in the paperSubstantial claim - Findingown data
Among rejected hyperplastic CADx outputs, 12.1% were correctly reclassified as SSLs.
“Similarly, although 43.6% of rejected hyperplastic CADx outputs were correctly reclassified as adenomas (31.5%; 95% CI, 24.3-39.7) or SSLs (12.1%; 95% CI, 7.5-18.7), there were 41.6% (62/149; 95% CI, 33.7-50.0) that were incorrectly reassigned as SSLs.”
Quote in the paperNot a substantial claim - Findingown data
CADx tends to predict hyperplastic at similar rates in the proximal (N = 120) and distal colon (N = 128).
“Furthermore, CADx tends to predict hyperplastic at similar rates in the proximal (N = 120) and distal colon (N = 128), but the final diagnoses are more frequently hyperplastic in the distal colon than proximal colon (N = 59 + 14 vs N = 9 + 17, respectively).”
Quote in the paperSubstantial claim - Findingown data
In the proximal colon, the proportion of correctly accepted CADx neoplastic diagnoses (91.1%) was higher than the proportion of correctly accepted CADx hyperplastic diagnoses (34.6%), P < .001.
“Looking at polyps located in the proximal colon, similar results were obtained, with a proportion of correctly accepted CADx neoplastic diagnoses (91.1%; 95% CI, 87.2-93.9, Figure S3) higher than the proportion of correctly accepted CADx hyperplastic diagnoses (34.6%; 95% CI, 17.9-55.6, Figure S4), P < .001.”
Quote in the paperNot a substantial claim
- Findingown data
Endoscopists accepted incorrect CADx classifications more frequently when the incorrect CADx classification was neoplastic compared to hyperplastic (64.9% vs 26.5%).
“Endoscopists accepted incorrect CADx classifications more frequently when these were neoplastic compared to hyperplastic (64.9% vs 26.5%).”
Quote in the paperSubstantial claim - Findingown data
Upon rejecting correct CADx neoplastic predictions, endoscopists misclassified 23.4% as sessile serrated lesions (SSLs).
“Upon rejecting correct CADx neoplastic predictions, endoscopists misclassified 23.4% as sessile serrated lesions (SSLs).”
Quote in the paperNot a substantial claim - Findingown data
Final diagnoses are more frequently hyperplastic in the distal colon than in the proximal colon.
“Furthermore, CADx tends to predict hyperplastic at similar rates in the proximal (N = 120) and distal colon (N = 128), but the final diagnoses are more frequently hyperplastic in the distal colon than proximal colon (N = 59 + 14 vs N = 9 + 17, respectively).”
Quote in the paperSubstantial claim - Findingown data
Among rejected hyperplastic CADx outputs, 31.5% were correctly reclassified as adenomas.
“Similarly, although 43.6% of rejected hyperplastic CADx outputs were correctly reclassified as adenomas (31.5%; 95% CI, 24.3-39.7) or SSLs (12.1%; 95% CI, 7.5-18.7), there were 41.6% (62/149; 95% CI, 33.7-50.0) that were incorrectly reassigned as SSLs.”
Quote in the paperNot a substantial claim - Findingown data
29.5% of rejected correct CADx hyperplastic predictions were incorrectly reassigned as SSLs.
“Similarly, 29.5% of rejected correct CADx hyperplastic predictions were incorrectly reassigned as SSLs.”
Quote in the paperNot a substantial claim - Findingown data
Accepted CADx diagnoses were correct for 89.1% of neoplastic CADx predictions.
“Accepted CADx diagnoses were correct for 89.1% of neoplastic and 68.7% of hyperplastic CADx predictions.”
Quote in the paperSubstantial claim - Findingown data
Among rejected neoplastic predictions (N = 47), endoscopists incorrectly reassigned 53.2% as SSLs.
“Regarding post-rejection reclassifications, among rejected neoplastic predictions (N = 47), endoscopists incorrectly reassigned 53.2% (25/47; 95% CI, 38.2-67.6) as SSLs.”
Quote in the paperSubstantial claim - Findingown data
In the distal colon, the proportions of correctly accepted CADx neoplastic (84.8%) and hyperplastic (80.8%) diagnoses were not statistically significantly different (P = .577).
“In the distal colon, the 2 proportions were not found to be statistically, significantly different (84.8%; 95% CI, 77.7 to 90.0 vs 80.8%; 95% CI, 69.6 to 88.8, respectively; P = .577).”
Quote in the paperNot a substantial claim - Findingown data
Among rejected hyperplastic CADx outputs, 41.6% were incorrectly reassigned as SSLs.
“Similarly, although 43.6% of rejected hyperplastic CADx outputs were correctly reclassified as adenomas (31.5%; 95% CI, 24.3-39.7) or SSLs (12.1%; 95% CI, 7.5-18.7), there were 41.6% (62/149; 95% CI, 33.7-50.0) that were incorrectly reassigned as SSLs.”
Quote in the paperSubstantial claim - Findingown data
Accepted CADx diagnoses were correct for 68.7% of hyperplastic CADx predictions.
“Accepted CADx diagnoses were correct for 89.1% of neoplastic and 68.7% of hyperplastic CADx predictions.”
Quote in the paperSubstantial claim - Premisereasoning
CADx demands greater expertise from endoscopists than CADe because endoscopists must use knowledge of polyp histology and clinical judgement to accept or reject AI-based polyp classifications.
“In contrast, CADx demands greater expertise from endoscopists, who must employ knowledge of polyp histology based on surface features and clinical judgement to accept or reject the AI-based polyp classifications.”
Quote in the paperSubstantial claim - Premisecitation
The clinical utility of AI-generated CADx predictions depends on the endoscopist's ability to interpret and appropriately act upon them.
“The clinical utility of these AI-generated predictions ultimately depends on the endoscopist’s ability to interpret and appropriately act upon them.4”
Quote in the paperSubstantial claim0 of 1 sources checked- cites [4] Oleksiw, 2025, Artificial intelligence and endoscopist diagnostic agreement as a framework for colorectal polyp optical diagnosis implementation, Clin Gastroenterol Hepatol source
- Interpretationown data
Endoscopists frequently override correct CADx hyperplastic diagnoses, mistakenly reclassifying these lesions as SSLs.
“We found that endoscopists frequently override correct CADx hyperplastic diagnoses, mistakenly reclassifying these lesions as SSLs.”
Quote in the paperSubstantial claim - Premisecitation
Previous studies show that diagnostic accuracy for CADx decreased for diminutive polyps <5 mm compared to small 6-9 mm and larger polyps >10 mm.
“Previous studies show that diagnostic accuracy for CADx also decreased for diminutive polyps <5 mm compared to small 6-9mm and larger polyps > 10 mm.14”
Quote in the paperSubstantial claim0 of 1 sources checked- cites [14] Li, 2023, Real-world validation of a computer-aided diagnosis system for prediction of polyp histology in colonoscopy: a prospective multicenter study, Am J Gastroenterol source
- Interpretationcitation
The inclusion of smaller polyps in the study may explain the variability in final diagnostic accuracy because accuracy of both endoscopists and CADx declines when polyp size decreases.
“The inclusion of smaller polyps in our study may explain the variability in final diagnostic accuracy since accuracy of both endoscopists and CADx declines when polyps size decreases.”
Quote in the paperSubstantial claim - Interpretationown data
Endoscopists are more prone to accept hyperplastic prediction in the distal colon.
“Thus, endoscopists are more prone to accept hyperplastic prediction in the distal colon.”
Quote in the paperSubstantial claim - Interpretationown data
This study is among the first to highlight such systematic diagnostic bias towards neoplastic polyps (adenoma or SSL) in real-world CADx implementation.
“This study is among the first to highlight such systematic diagnostic bias towards neoplastic polyps (adenoma or SSL) in real-world CADx implementation, a phenomenon with important clinical and potential economic implications.”
Quote in the paperSubstantial claim - Study design and conductreasoning
The mismatch between binary CADx outputs and the endoscopist three-category scheme introduced a confounding factor in the study.
“Currently available CADx systems provide binary polyp classification as neoplastic or hyperplastic, whereas endoscopists routinely employ a three-category scheme (adenoma, hyperplastic, SSL), which introduced a confounding factor in our study.”
Quote in the paperSubstantial claim - Interpretationreasoning
Many apparent reclassification errors may not actually reflect human misjudgment, but rather a semantic misalignment between the CADx system's output and the endoscopists' diagnostic categories.
“Many apparent reclassification errors may not actually reflect human misjudgment, but rather a semantic misalignment between the CADx system’s output and the endoscopists’ diagnostic categories.”
Quote in the paperSubstantial claim - Interpretationreasoning
The observed bias towards classifying benign lesions as neoplastic suggests CADx-assisted diagnoses are generally safe.
“From a clinical risk perspective, the observed bias towards classifying benign lesions as neoplastic suggests CADx-assisted diagnoses are generally safe.”
Quote in the paperSubstantial claim - Interpretationown data
Endoscopists are more likely to label polyps as hyperplastic when located in the distal colon.
“This suggests that endoscopists are more likely to label polyps as hyperplastic when located in the distal colon.”
Quote in the paperSubstantial claim
- May combine two claims. 7
- Classified as probably not a substantial claim. 1
May combine two claims.
Interpretationown dataEndoscopists exhibit a diagnostic bias toward neoplasia, particularly misclassifying hyperplastic polyps as SSLs.
“Endoscopists exhibit a diagnostic bias toward neoplasia, particularly misclassifying hyperplastic polyps as SSLs.”
- Also stated in: § Discussion · boosted
“In conclusion, endoscopists demonstrate a pronounced diagnostic bias towards neoplastic lesions, frequently misclassifying hyperplastic polyps as SSLs.”
Quote in the paperSubstantial claim- Also stated in: § Discussion · boosted
May combine two claims.
Study design and conductreasoningThe lack of data concerning pathologists and the unavailability of Kappa among pathologists interpreting hyperplastic polyps and SSLs could create a confounding factor because histopathology is the gold standard in the study.
“Another limitation to this study is the lack of data concerning the pathologists. The Kappa among pathologists interpreting hyperplastic polyps and SSLs is not available. Since the gold standard in this study is the histopathology, this could create a confounding factor.”
Quote in the paperNot a substantial claimMay combine two claims.
InterpretationreasoningThe diagnostic bias toward neoplasia may risk unnecessary surveillance colonoscopies and increase costs.
“While this may enhance safety, it may risk unnecessary surveillance colonoscopies and increase costs.”
Quote in the paperSubstantial claimMay combine two claims.
Premisenone givenAdenoma identification typically relies on clearly defined morphological features such as vessel and pit patterns, while SSL diagnosis often relies on exclusion criteria, resulting in diagnostic uncertainty and overdiagnosis.
“While adenoma identification typically relies on clearly defined morphological features such as vessel and pit patterns, SSL diagnosis often relies on exclusion criteria, thus resulting in diagnostic uncertainty and overdiagnosis.”
Quote in the paperSubstantial claimMay combine two claims.
PremisecitationIn clinical practice, small polyps <10 mm are more frequently encountered and are more difficult for endoscopists to accurately diagnose.
“In clinical practice, small polyps <10 mm are more frequently encountered and are more difficult for endoscopist to accurately diagnose.14”
Quote in the paperSubstantial claim0 of 1 sources checked- cites [14] Li, 2023, Real-world validation of a computer-aided diagnosis system for prediction of polyp histology in colonoscopy: a prospective multicenter study, Am J Gastroenterol source
May combine two claims.
Interpretationown dataOur results extend previous work by quantifying the consequences of overrides: significant overdiagnosis of SSL/Ps and consequent inflation of surveillance needs.
“Our results extend this work by quantifying the consequences of such overrides: significant overdiagnosis of SSL/Ps and consequent inflation of surveillance needs.”
Quote in the paperSubstantial claimClassified as probably not a substantial claim.
InterpretationreasoningThe diagnostic bias toward neoplasia may enhance safety.
“While this may enhance safety, it may risk unnecessary surveillance colonoscopies and increase costs.”
Quote in the paperSubstantial claimMay combine two claims.
Premisenone givenCurrently available CADx systems provide binary polyp classification as neoplastic or hyperplastic, whereas endoscopists routinely employ a three-category scheme (adenoma, hyperplastic, SSL).
“Currently available CADx systems provide binary polyp classification as neoplastic or hyperplastic, whereas endoscopists routinely employ a three-category scheme (adenoma, hyperplastic, SSL), which introduced a confounding factor in our study.”
Quote in the paperSubstantial claim
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
Each contradiction between two passages listed here was confirmed by an independent check.
Cohort size differs between abstract and results
MinorThe abstract reports 754 included polyps, while the Results section and Table 1 report 753.
“Our cohort included 754 polyps <10 mm from 367 patients.”
“753 polyps found in 367 patients (mean age 66, 46.3% female) were included in our secondary analysis”
Cites
Computer-aided detection-assisted colonoscopy: classification and relevance of false positives
Hassan, C., Badalamenti, M., Maselli, R., Correale, L., Iannone, A., Radaelli, F., Rondonotti, E., Ferrara, E., Spadaccini, M., Alkandari, A., et al. (2020) · Gastrointestinal Endoscopy · doi:10.1016/j.gie.2020.06.021 · Cited by 81 · see citing works
Self-citationArtificial intelligence-assisted colonoscopy: A review of current state of practice and research
Taghiakbari, M., Mori, Y., & Renteln, D. A. V. (2021) · World Journal of Gastroenterology · doi:10.3748/wjg.v27.i47.8103 · Cited by 71 · see citing works
Parasa, S., Berzin, T., Leggett, C., Gross, S., Repici, A., Ahmad, O. F., Chiang, A., Coelho-Prabhu, N., Cohen, J., Dekker, E., et al. (2025) · Gastrointestinal Endoscopy · doi:10.1016/j.gie.2023.12.003 · Cited by 64 · see citing works
Self-citationMori, Y., East, J. E., Hassan, C., Halvorsen, N., Berzin, T. M., Byrne, M. F., Renteln, D. A. V., Hewett, D. G., Repici, A., Ramchandani, M. J., et al. (2023) · Digestive Endoscopy · doi:10.1111/den.14531 · Cited by 62 · see citing works
Self-citationLi, J. W., Wu, C. C. H., Lee, J. W. J., Liang, R., Soon, G. S. T., Wang, L. M., Koh, X. H., Koh, C. J., Chew, W. D., Lin, K. W., et al. (2023) · American Journal of Gastroenterology · doi:10.14309/ajg.0000000000002282 · Cited by 61 · see citing works
Current status and limitations of artificial intelligence in colonoscopy
Hann, A., Troya, J., & Fitting, D. (2021) · United European Gastroenterology Journal · doi:10.1002/ueg2.12108 · Cited by 55 · see citing works
Simplifying Resect and Discard Strategies for Real-Time Assessment of Diminutive Colorectal Polyps
Renteln, D. A. V., Kaltenbach, T. R., Rastogi, A. N., Anderson, J. C., Rösch, T., Soetikno, R. M., & Pohl, H. (2017) · Clinical Gastroenterology and Hepatology · doi:10.1016/j.cgh.2017.11.036 · Cited by 42 · see citing works
Self-citationvan der Zander, Q. E. W., Roumans, R., Kusters, C. H. J., Dehghani, N., Masclee, A. A. M., de With, P. H. N., van der Sommen, F., Snijders, C. C. P., & Schoon, E. J. (2024) · Gastrointestinal Endoscopy · doi:10.1016/j.gie.2024.06.029 · Cited by 27 · see citing works
Rizkala, T., Hassan, C., Mori, Y., Spadaccini, M., Antonelli, G., Dekker, E., Houwen, B. B. S. L., Pech, O., Baumer, S., Rondonotti, E., et al. (2024) · Clinical Gastroenterology and Hepatology · doi:10.1016/j.cgh.2024.08.021 · Cited by 20 · see citing works
Self-citationPragmatic Resect and Discard Implementation Using Computer-Assisted Optical Polyp Diagnosis
Taghiakbari, M., Rex, D. K., Pohl, H., Djinbachian, R., Huang, F., Hassan, C., & Renteln, D. A. V. (2024) · Gastroenterology · doi:10.1053/j.gastro.2024.08.037 · Cited by 14 · see citing works
Self-citationDjinbachian, R., Taghiakbari, M., Barkun, A. N., Medawar, E., Alj, A., Sidani, S., Kiow, J. L. C., Panzini, B., Bouin, M., Lasfar, D., et al. (2024) · Surgical Endoscopy · doi:10.1007/s00464-024-11466-7 · Cited by 10 · see citing works
Self-citationHalvorsen, N., Barua, I., Bretthauer, M., Mori, Y., Kudo, S.-E., Gulati, S., Misawa, M., Mori, K., Hayee, B., Olabintan, O., et al. (2025) · Endoscopy · doi:10.1055/a-2532-9282 · Cited by 9 · see citing works
Self-citationByrne, M. F., Renteln, D. A. V., & Barkun, A. N. (2023) · Gastroenterology · doi:10.1053/j.gastro.2023.01.003 · Cited by 8 · see citing works
Optical Polyp Diagnosis in the Era of Artificial Intelligence
Djinbachian, R., Rex, D. K., & Renteln, D. A. V. (2024) · The American Journal of Gastroenterology · doi:10.14309/ajg.0000000000003195 · Cited by 5 · see citing works
Self-citationOleksiw, M., Djinbachian, R., & Von Renteln, D. (2025) · Endoscopy · doi:10.1055/s-0045-1805444 · Cited by 2 · see citing works
Self-citation
Flagged first, then by citations
Research works printed but not linked: 1
- R Core Team, 2024, R: A Language and Environment for Statistical Computing
Cited by
Still being checked.
- Authors
- Dubois, D., Jalal, Y., Pohl, H., Rex, D. K., Hassan, C., Djinbachian, R., Oleksiw, M., Michal, V., & von Renteln, D.
- Affiliations
- Université de Montréal (CA)
- United States Department of Veterans Affairs (US)
- Indiana University School of Medicine (US)
- IRCCS Humanitas Research Hospital (IT)
- Centre Hospitalier de l’Université de Montréal (CA)
- University of Montreal Hospital Research Center (CRCHUM) , Montreal, Quebec, H2X0A9,
- ORCID
- Venue
- Journal of the Canadian Association of Gastroenterology
- Publisher
- Oxford University Press (OUP)
- Published
- 2026-04-29
- Volume
- 9
- Issue
- 4
- Pages
- 204-209
- ISSN
- 2515-2084 (ISSN-L)
- 2515-2092
- PMC
- PMC13436811
- OpenAlex
- W7159965148
- Cited by
- 0
- References
- 15
- FWCI
- 0.00
- Type
- journal article
- Language
- en
- Open access
- gold
- doi.org
- Licence
- Locations
- Journal of the Canadian Association of Gastroenterology (publishedVersion, cc-by-nc-nd)
- PubMed (publishedVersion)
- PubMed Central (publishedVersion, cc-by-nc-nd)
- Funders
- Fonds de Recherche du Québec Santé
- Topics
- Health Sciences › Medicine › Oncology › Colorectal Cancer Screening and Detection
- Physical Sciences › Computer Science › Artificial Intelligence › AI in cancer detection
- Health Sciences › Medicine › Health Informatics › Artificial Intelligence in Healthcare and Education
- Keywords
- computer-aided diagnosis, colorectal polyps, human-machine interaction, sessile serrated lesions, colonoscopy, histopathology validation
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Claims read by deepseek-v4.1-flash under definition version 3 (extract-v3) on 2026-11-03.
Classified by jev-1.13-20260917 (gold-guideline-v0.2) on 2026-11-03.
Cited sources are checked by gpt-6-sol.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
