ERCPMP-Gx: Endoscopic Image and Video Dataset for Morphological, Histopathological, and Genomic Characterization of Colorectal Polyposis
Zahra Ghaffari, Massih Bahar, Mojgan Forootan, Ali Darvishi, Hamidreza Bolhasani
cs.CV · cs.AI
Quick verdict
First public patient-level endoscopy-pathology-genetics polyposis dataset: novel and unusually candid, but patient counts unreported and no validation results.
Summary
This data descriptor introduces ERCPMP-Gx, a publicly released multimodal dataset linking endoscopic imaging, histopathology, and germline genetic findings at the patient level for hereditary colorectal polyposis syndromes. Collected between 2024 and 2026 at an Iranian referral center with Olympus EVIS X1 equipment across white-light, NBI, and magnifying NBI modes, the resource comprises 160 still images plus video clips covering FAP, PJS, JPS, and ganglioneuroma syndrome (~80% of cases, clinically and/or genetically confirmed) and non-hereditary mimics such as inflammatory and pneumatosis intestinalis lesions (~20%). Ground truth followed a four-step pipeline: targeted biopsy, independent review by two GI pathologists, multi-expert endoscopic annotation using Paris, Pit Pattern, and JNET schemes, and consensus adjudication. The authors honestly state that no train/test split, no inter-rater reliability statistics, and no baseline model results are provided, and that real-time genotype prediction remains an unproven future direction. The main weaknesses are incomplete cohort accounting — the number of patients is never reported, Table 2 leaves several syndrome columns blank, and video counts are absent — plus internal inconsistencies in equipment and syndrome lists. As a pilot resource for proof-of-concept research it is genuinely novel and unusually candid.
Review complete
Assessment at a glance
5
risk flags identified
● high 0
● medium 2
● low 3
4 strengths · 4 weaknesses · 5 author questions
AI text signal = Binoculars zero-shot detector score over 20 sampled passages (Fast-DetectGPT co-reported) — absence does not prove human authorship, and a flag does not imply misconduct.
Levels reflect verification complexity — not a quality score. Evolve never rates your paper.
Key claims under review
- ERCPMP-Gx is the first publicly available dataset integrating colonoscopic imaging, histopathology, and germline genetics at the patient level for hereditary polyposis syndromes (Sections 1.2, Table 1).
- The released resource comprises 160 endoscopic images plus videos, ~80% clinically and/or genetically confirmed hereditary cases (FAP, PJS, JPS, GNS) and ~20% non-hereditary morphologic mimics.
- Ground truth was established through a four-step process culminating in dual pathologist review and consensus adjudication of discordant cases (Section 4.1).
- Real-time genotype prediction is presented strictly as a potential future research direction enabled by the resource, not as a demonstrated capability (Conclusion).
Strengths
Genuine novelty in patient-level multimodal linkage
Table 1 supports the claim that prior resources (HyperKvasir: 110,079 images, no hereditary focus; ERCPMP 2024: 217 WLE images, no germline data; PICCOLO: ex vivo histopathology only) lack hereditary focus or genomic linkage; ERCPMP-Gx links WLE/NBI/M-NBI imagery to histopathology and germline findings per patient (Section 1.2).
Rigorous four-step ground-truth pipeline
Section 4.1: targeted biopsy/polypectomy, independent assessment by two board-certified GI pathologists, multi-gastroenterologist annotation under Paris/Pit Pattern/JNET schemes, consensus adjudication; Section 2 separates endoscopic phenotype from diagnostic outcome, and metadata explicitly distinguish histopathological confirmation from optical impression — an anti-circularity design most polyp datasets lack.
Unusual scientific candor
Practical release hygiene
Weaknesses
Incomplete cohort accounting
Internal inconsistencies
No quantitative validation of any kind
Severe scale and bias constraints
Full Review Report
$1.99 · one-timeUnlock the complete preliminary analysis Evolve produced for this paper.
Report preview
Scope & Claims
ERCPMP-Gx (arXiv 2609.20815) is a data descriptor for a multimodal endoscopic dataset targeting hereditary colorectal polyposis syndromes. The central claim, stated in Sections 1.1-1.2 and supported by Table 1, is that this is the first publicly available resource linking colonoscopic imaging (WLE, NBI, M-NBI, near focus), histopathological confirmation, and clinically reported germline findings at the patient level, enabling joint phenotype-genotype modeling for FAP, PJS, JPS, and ganglioneuroma syndrome. The cohort was assembled 2024-2026 through a single Iranian referral pathway (FamCan institute to an advanced gastroenterology ward, Shahid Beheshti University) and comprises 160 still images plus video clips: roughly 80% hereditary (PG) and 20% non-hereditary mimics (Non-PG) including inflammatory/IBD-like polyps, P-SRUS polyps, and pneumatosis intestinalis. Notably, the final paragraph of the Conclusion explicitly disclaims that the descriptor does not demonstrate real-time genotype prediction; that is positioned strictly as future work. This candor shapes a fair reading: the claim surface is the dataset itself, not model results.
The remaining sections cover Methodology Assessment, Statistical Evidence, Results & Reproducibility, Limitations & Risks, and Verdict & Recommendations.
- All 4 strengths & 4 weaknesses, analyzed in detail
- 5 risk flags with severity, section & reviewer questions
- Methods & statistical evidence assessment
- 5 questions a peer reviewer would ask
- Full six-dimension evaluation report & recommendations
One-time purchase · this paper only
One-time purchase, this paper only. AI-generated preliminary analysis — not peer review.
Frequently asked questions
What is the ERCPMP-Gx dataset?
ERCPMP-Gx is a publicly available multimodal dataset of endoscopic images and video clips from patients with hereditary colorectal polyposis syndromes (FAP, PJS, JPS, ganglioneuroma syndrome) plus non-hereditary lookalike lesions. Each record is linked where available to histopathology and germline genetic findings at the patient level. It contains 160 still images plus videos and is hosted on Mendeley under a CC BY-NC 4.0 license.
How many patients are in the ERCPMP-Gx dataset?
The paper does not report a patient count. It states the dataset contains 160 endoscopic still images plus video clips, with roughly 80% hereditary and 20% non-hereditary cases. To determine the actual number of patients, you would need to inspect the metadata file in the Mendeley deposit itself.
Can this dataset train an AI to predict genetic mutations from colonoscopy images?
No capability is demonstrated. The authors explicitly state in the paper's conclusion that real-time genotype prediction from endoscopic imaging is only a potential future research direction, and the paper reports no model results. Any such claim would require new modeling studies with large cohorts and external validation.
Which syndromes and imaging modalities does ERCPMP-Gx cover?
Hereditary cases cover familial adenomatous polyposis (FAP), Peutz-Jeghers syndrome (PJS), juvenile polyposis syndrome (JPS), and ganglioneuroma syndrome (GNS), with Cowden syndrome listed in the inclusion criteria; the non-hereditary comparator group includes inflammatory, P-SRUS, and pneumatosis intestinalis lesions. Imaging uses white-light endoscopy, narrow-band imaging (NBI), magnifying NBI, and NBI with near focus on Olympus EVIS X1 equipment.
About this review
This is an AI-generated preliminary review by Evolve.Pub, produced from the openly available text of the paper. It is not peer review and does not represent the authors' views. The canonical record of the paper is at the publisher link above.
- Table 1 arrived heavily garbled and Table 2 shows dash cells that could be extraction artifacts rather than true blanks in the PDF — the cohort-accounting criticism should be re-checked against the published tables.
- Figures 1-3 (sample images, metadata screenshot, graphical abstract) could not be inspected from the extracted text.
- The provided abstract is truncated mid-sentence ('linked, w...'), so the full abstract may contain additional detail.
- Actual Mendeley repository contents (video counts, metadata fields, granularity of genetic data) were not verifiable from the text.
Go deeper on this paper
Wrote a paper? Stress-test it before reviewers do.
Upload your manuscript for the same structured scrutiny — reasoning, statistics, evidence, figures, and reproducibility.