GastroNet-5M: A Multicenter Dataset for Developing Foundation Models in Gastrointestinal Endoscopy
This study presents GastroNet-5M, a dataset of ∼5 million endoscopic images. Pretraining endoscopic deep learning systems with GastroNet-5M improves diagnostic accuracy, reduces the need for scarce application-specific endoscopic imagery and annotations, and increases their robustness to the inevitable data heterogeneity in clinical practice. This may significantly accelerate development and implementation of endoscopic AI systems. GastroNet-5M is publicly available for scientific use.
Authors
- Maarten R. Struyvenberg
- Kiki Fockens
- Tessa Jaspers (ORCID: https://orcid.org/0000-0002-3939-1403)
- Peter H. N. de With (ORCID: https://orcid.org/0000-0002-7639-7716)
- Joost van der Putten (ORCID: https://orcid.org/0000-0002-0915-9048)
- Carolus H. J. Kusters (ORCID: https://orcid.org/0009-0004-3114-3888)
- J Bergman
- Fons van der Sommen (ORCID: https://orcid.org/0000-0002-3593-2356)
- Tim Boers (ORCID: https://orcid.org/0000-0002-5309-1941)
- Raf Bisschops (ORCID: https://orcid.org/0000-0002-9994-8226)
- A J de Groof
- F C Slooter
- J B Jukema
- Rixta A. H. van Eijck van Heslinga
- M R Jong
Institutions
- Amsterdam Neuroscience (NL)
- Amsterdam University Medical Centers (NL)
- Eindhoven University of Technology (NL)
- University of Amsterdam (NL)
- KU Leuven (BE)
Publication Details
- Journal
- Gastroenterology
- Published
- 2025-07-30
- DOI
- https://doi.org/10.1053/j.gastro.2025.07.030
- Citations
- 8
- Primary Topic
- Colorectal Cancer Screening and Detection
- Type
- article
- Field-Weighted Citation Impact
- 8.36
Funders
- Nederlandse Organisatie voor Wetenschappelijk Onderzoek
- KWF Kankerbestrijding