Accuracy of pathology and computer-assisted optical diagnosis of diminutive colorectal polyps based on expert image and video audit as the reference standard
Bibliographic record
Abstract
Aims Optical polyp diagnosis (OD) is less resource-intensive and costly than pathology. Moreover, the increasing use of OD and Computer-assisted OD (CADx) has in recent years shown that pathological diagnosis can be incorrect. Several studies have shown that polyps diagnosed in pathology as mucosal polyps are likely misdiagnoses due to errors in resection, retrieval, embedding, sectioning, or pathology interpretation of the lesions. We were interested in comparing pathology-based diagnosis to CADx using agreement between 3 expert endoscopists as the gold standard for diagnosis. Methods We conducted a prospective study in which three experts (DKR, CH, HP) evaluated all polyps of a large prospective cohort undergoing CADx assisted OD based on image and video material. We assumed that when three experts agree with high confidence on an unanimous polyp diagnosis in a blinded review, that this diagnostic agreement constitutes a hierarchical higher ground-truth than pathology, OD or CADx. Experts were blinded to the initial CADx-assisted OD, pathology result, and each other’s OD. Primary outcome was diagnostic accuracy CADx-assisted OD compared to pathology for diminutive (≤ 5 mm) colorectal polyps using expert image and video audit as the reference standard. We hypothesized that the accuracy of CADx-assisted OD is non-inferior to pathological interpretation for diminutive polyps. Results Among 511 patients recruited in the study, 523 diminutive colorectal polyps were identified. 487 polyps were reviewed by the expert endoscopists. The experts agreed with high confidence on an unanimous polyp diagnosis for 225 polyps. Based on unanimous polyp diagnosis between three experts as reference standard CADx-assisted OD was accurate in 205 of 225 polyps (91.1%) (95% CI: (87.4, 94.8)) and histopathological interpretation was accurate in 173 polyps (76.9%) (95% CI: (71.4, 82.4)), (paired p-value<0.001). CADx-assisted OD agreed with pathology in 165 polyps (73.3%) (95% CI: (67.6, 79.1)). Pathology reported 21 normal tissue specimens among the 225 polyps (9.33%), 7 of which were unanimously diagnosed as hyperplastic polyps by the experts and CADx. The remaining 14 of 21 histologically diagnosed normal mucosa were diagnosed as high-confidence adenomas by the experts. CADx agreed with the experts’ OD in 13 of 14 adenomas (92.9%). Conclusions This study demonstrated a statistically significant higher diagnostic accuracy when CADx-assisted OD is used compared to the standard histopathological assessment for diminutive colorectal polyps. It provides a proof of concept that CADx-assisted OD might indeed outperform pathology in diagnostic accuracy for diminutive polyps. Similar to previous studies, we found that a considerable number of polyps are incorrectly diagnosed as normal mucosa in pathology. Validity of our results is supported by a growing body of evidence showing that reevaluation in pathology of such discrepant findings often lead to revision of the initial pathology diagnosis. Publication History Article published online: 27 March 2025 © 2025. European Society of Gastrointestinal Endoscopy. All rights reserved. Georg Thieme Verlag KG Oswald-Hesse-Straße 50, 70469 Stuttgart, Germany
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
The three-model screen
all 5,600 screened works →All three models called this out of scope.
Diagnostic accuracy study of optical diagnosis of colorectal polyps; abstract absent, and the reference standard question is clinical rather than metaresearch.
It evaluates diagnostic accuracy for colorectal polyps.
Title is a clinical diagnostic-accuracy study of polyp assessment; empty abstract but object is clinical.
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.005 | 0.019 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.001 |
| Bibliometrics | 0.003 | 0.001 |
| Science and technology studies | 0.000 | 0.001 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.001 | 0.000 |
| Insufficient payload (model declined to judge) | 0.001 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".