Introduction
Artificial intelligence (AI) could improve endoscopic diagnosis and outcome assessment in gastrointestinal (GI) diseases. Robust AI models require large, labeled datasets with high-resolution endoscopic videos. Many existing endoscopic databases rely on retrospective sampling, isolated images, or lack clinical context, limiting their utility for AI system development.1-4 We aimed to develop an endoscopy video databank that couples full-length recording with capture of structured clinical, procedural, and pathology data.
Aims & Methods
A prospective video databank was established at the Montreal University Hospital Center (NCT06822616) with enrollment of patients ≥ 18 years old undergoing upper (EGD) or lower (colonoscopy) endoscopic procedures since 2022. All patients provided informed consent for data collection and subsequent use in AI system development. The databank follows FAIR principles: Findable, Accessible, Interoperable, and Reusable.5 Recordings were de-identified from acquisition, using study identifiers without protected health information, with all technical specifications optimized for AI training (full-length, full-frame videos in 1920x1080-pixel resolution at 60fps, 10-bit color depth). Trained research staff systematically documented and timestamped anatomical landmarks, all identified lesions, including endoscopist optical diagnoses, interventions, disease-specific scores, AI outputs, and endoscopy quality metrics in structured case report forms (CRFs) during the procedure.
Results
1As of July 2025, 6,272 patients (mean 59.0 years, 52.1% female) have been prospectively enrolled for a total of 7,591 procedures (5,605(73.8%) colonoscopies, 1,986(26.2%) EGDs, Table 1) performed by 48 different endoscopists. Up to 747 and 455 structured datapoints were documented for each enrolled colonoscopy and EGD, respectively, capturing substantial diversity across inflammatory, neoplastic, and structural disease categories. Each polyp identified and biopsy performed were documented with up to 48 and 46 variables each. Disease-specific datapoints included up to 169 datapoints specifically for inflammatory bowel disease. Pathologic findings (including polyps, inflammation, dysplasia, or malignancy) found in 6355 (83.7%) records were documented in the cohort, of which 6987 colorectal polyps across 2981 procedures. Data enabled calculation of multiple endoscopic quality metrics, with a mean adenoma detection rate of 38.9% [95% CI: 37.2-40.7] and a mean withdrawal time of 12.1 minutes across the colonoscopy cohort.
Table 1. Overview of cohort characteristics: prevalence of procedural indications, endoscopic and histological findings
| Colonoscopies | N=5,605 | EGD | N=1,986 |
| Main procedure indications, n (%) | Surveillance | 1,931 (34.4) | Dyspepsia | 306 (15.6) |
| Diagnostic | 760 (13.6) | Barrett’s esophagus | 229 (11.5) |
| Screening | 642 (11.4)
| Anemia | 198 (10.0) |
| Crohn’s disease | 543 (9.7)
| Dysphagia | 151 (7.6) |
| Ulcerative colitis | 459 (8.2)
| Surveillance | 126 (6.3) |
| Endoscopic and histological findings, n | Colorectal polyps | 6987 | Barrett’s esophagus | 280 |
| Colorectal Cancer | 48 | Reflux esophagitis | 112 |
| Active Crohn’s disease | 178 | Ulcer | 19 |
| Active ulcerative colitis | 198 | Gastritis | 331 |
Conclusion
This large, prospective, FAIR-compliant video databank of >7,500 procedures enrolled at our tertiary care center provides a scalable foundation for AI development, validation, and clinical outcomes research for GI endoscopy. Ongoing multi-center expansion and integration of digital pathology slides, microbiome and multi-omics analyses will enable future projects combining endoscopic video data with precision medicine approaches.
References
1. Steinhäuser JL, Berzin TM, Geissler ME, et al. Implementing endoscopy video recording in routine clinical practice: strategies from three tertiary care centers. Endosc Int Open 2025;13:a25923338.
2. Biffi C, Antonelli G, Bernhofer S, et al. REAL-Colon: a dataset for developing real-world AI applications in colonoscopy. Sci Data 2024;11:539.
3. Jong MR, Boers TGW, Fockens KN, et al. GastroNet-5M: a multicenter dataset for developing foundation models in gastrointestinal endoscopy. Gastroenterology 2026;170:174-87.
4. Bravo D, Frias J, Vera F, et al. GastroHUN an endoscopy dataset of complete systematic screening protocol for the stomach. Sci Data 2025;12:102.
5. Wilkinson MD, Dumontier M, Aalsbersberg IJ, et al. The FAIR Guiding Principles for scientific data management and stewardship. Sci Data 2016;3:160018.
Disclosure
Daniel von Renteln has received research funding from ERBE Elektromedizin GmbH, Boston Scientific, Medtronic, Ventage, Magentiq eye, Dova Health, Olympus, Fujifilm, and Cosmo IMD, and has received consultant or speaker fees from Medtronic, Fujifilm, Olympus, Dova Health, Magentiq eye, Boston Scientific Inc., and ERBE Elektromedizin GmbH. Roupen Djinbachian has received speaker fees from Fujifilm. The remaining authors declare that they have no conflict of interest.