ERCPMP-Gx: Multimodal Endoscopic, Histopathological, and Genomic Dataset for Hereditary Polyposis Syndromes

Forum topic · 小凯 · 2026-09-19

Summary

ERCPMP-Gx is a publicly available multimodal dataset designed to support artificial intelligence applications in recognizing, characterizing, and classifying hereditary colorectal polyposis syndromes—precursor lesions to colorectal cancer linked to a broad spectrum of extracolonic tumors. Most examinations used the Olympus EVIS X1 system with white light endoscopy, narrow band imaging (NBI), magnified NBI, and NBI near-focus modes, yielding 160 images plus paired video clips. Around 80% of cases are clinically and/or genetically confirmed hereditary polyposis syndromes, including familial adenomatous polyposis (FAP), Peutz-Jeghers syndrome (PJS), juvenile polyposis syndrome (JPS), and ganglioneuromatosis syndrome (GNS); the remaining 20% are non-hereditary polyps and phenotypic mimics included for differential classification. Each record links standardized endoscopic annotations, representative histopathology, and germline findings where feasible, forming an AI-ready patient-level annotation framework. The dataset is available on Mendeley (https://doi.org/10.17632/nzyfc544bx.2), with updates hosted at https://databiox.com. arXiv: https://arxiv.org/abs/2609.20815.

Overview

Research field: Computer Vision Authors: Zahra Ghaffari, Massih Bahar, Mojgan Forootan, Ali Darvishi, Hamidreza Bolhasani Published: 2026-09-17 arXiv: 2609.20815

Key Points

Hereditary polyposis syndromes can be precursor lesions to colorectal cancer and are associated with a broad spectrum of extracolonic tumors. Early identification and accurate classification are essential for timely diagnosis, individualized patient management, and targeted surveillance of affected families.
Existing public endoscopic datasets focus on individual sporadic polyps; none links the polyposis phenotype to histopathology and germline findings at the patient level.
ERCPMP-Gx is an endoscopic, histopathological, and genomic dataset built to support AI in recognition, characterization, and classification of colorectal polyposis.
Most procedures used the Olympus EVIS X1 system, including white light endoscopy (WLE), narrow band imaging (NBI), magnified NBI (M-NBI), and NBI near-focus mode, producing 160 images with accompanying video clips.
~80% of cases are clinically and/or genetically confirmed hereditary polyposis syndromes (PG): familial adenomatous polyposis (FAP), Peutz-Jeghers syndrome (PJS), juvenile polyposis syndrome (JPS), and ganglioneuromatosis syndrome (GNS).
The remaining ~20% are non-hereditary polyps and morphologically overlapping polyposis-mimicking lesions (Non-PG), supporting differential classification.
Each published record links standardized endoscopic annotations, representative histopathology, and clinically reported germline findings where feasible—an AI-ready patient-level annotation framework.

ERCPMP-Gx: Multimodal Endoscopic, Histopathological, and Genomic Dataset for Hereditary Polyposis Syndromes

Summary

Overview

Key Points

Links

Tags