A team of researchers has released ERCPMP-Gx, a new dataset that connects colonoscopy images and videos with tissue pathology and inherited genetic test results for people with colorectal polyposis. The aim is to give artificial intelligence (AI) tools the combined information they need to recognize and classify hereditary polyposis syndromes during routine exams.

The dataset was assembled from procedures performed between 2024 and 2026 on patients referred to a familial cancer clinic. Most recordings were made with an Olympus EVIS X1 colonoscopy system and include white‑light endoscopy (WLE), narrow‑band imaging (NBI), magnifying NBI (M‑NBI), and NBI with near focus. In total the release contains 160 images plus accompanying video clips converted to PNG and MP4 formats. About 80% of cases are clinically and/or genetically confirmed hereditary syndromes such as familial adenomatous polyposis (FAP), Peutz‑Jeghers syndrome (PJS), juvenile polyposis syndrome (JPS), and ganglioneuroma syndrome (GNS). The remaining cases are non‑hereditary polyps or lesions that look similar, included to help AI tell them apart.

What makes ERCPMP-Gx different from earlier public collections is that each patient record, where available, links standardized endoscopic annotations to representative histopathology (microscope results) and reported germline genetic findings (inherited DNA changes). The authors present this as an AI‑ready, patient‑level annotation framework designed to train models that can jointly learn an endoscopic appearance and its underlying tissue and genetic causes. They suggest this could support real‑time AI that predicts pathology and possible genetic diagnoses while the doctor is performing a colonoscopy.

The dataset could matter because hereditary polyposis syndromes carry a high risk for colorectal cancer and for tumors in family members. Better, earlier recognition during colonoscopy can change how patients are monitored and how families are counseled. By combining images, pathology, and genetics, ERCPMP‑Gx aims to enable studies that go beyond simple polyp detection and toward phenotype‑to‑genotype prediction.