Copyright: ©Author(s) 2026. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution-NonCommercial (CC BY-NC 4.0) license. No commercial re-use. See permissions. Published by Baishideng Publishing Group Inc.
World J Gastroenterol. Dec 7, 2026; 32(45): 121242
Published online Dec 7, 2026. doi: 10.3748/wjg.121242
Published online Dec 7, 2026. doi: 10.3748/wjg.121242
Domain specific training with three-dimensional virtual colonoscopy images improves artificial intelligence based automated size prediction for large lesions
Dominik Schulz, Eva Holley, Sandra Nagl, Andreas Probst, Markus W Scheppach, Christoph Römmele, David Roser, Stephan Zellmer, Helmut Messmann, III. Medizinische Klinik, Universitätsklinikum Augsburg, Augsburg 86156, Germany
Alanna Ebigbo, Medizinische Klinik I, Universitätsklinikum St. Josef-Hospital, Bochum 44791, North Rhine-Westphalia, Germany
Olga Nichiporyk, Institut für Transfusionsmedizin und Hämostaseologie, Universitätsklinikum Augsburg, Augsburg 86156, Germany
Author contributions: Schulz D designed and conducted the research study, wrote the original draft; Ebigbo A, Holley E, Nagl S, Nichiporyk O, Probst A, and Zellmer S provided size estimations; Messmann H provided funding acquisition; Schulz D, Ebigbo A, Holley E, Nagl S, Nichiporyk O, Probst A, Scheppach MW, Römmele C, Roser D, Zellmer S, and Messmann H reviewed and edited the manuscript; all of the authors read and approved the final version of the manuscript to be published.
AI contribution statement: During the preparation of this work the main author used Gemini 3 in order to refine some of the paragraphs for proper journal english. After using this tool/service, the main author reviewed and edited the content as needed and takes full responsibility for the content of the published article.
Institutional review board statement: The study was reviewed and approved by the Ethics Committee, No. 25-4152-104.
Informed consent statement: The data used in this study were originally collected during routine clinical care, and the analysis was performed exclusively within the institution in compliance with applicable data protection regulations. According to Article 27(4) of the Bavarian Hospital Act (Bayerisches Krankenhausgesetz, BayKrG), individual patient consent is not required for internal retrospective analyses conducted for quality assurance or scientific evaluation within the hospital. Therefore, the study was carried out without obtaining separate informed consent from each patient.
Conflict-of-interest statement: Ebigbo A reports consultancy fees from Ambu and Olympus. Römmele C reports consultancy fees from Ambu Innovations and Ambu GmbH, and unpaid consultancy for Boston Scientific. Messmann H reports relationships with Ambu, Boston Scientific, Fujifilm, Hitachi, and Olympus; has received honoraria from Olympus, and hasreceived consultation fees from Ambu, Boston Scientific, and Olympus. Schulz D, Holley E, Nagl S, Nichiporyk O, Probst A, Scheppach MW, Roser D, and Zellmer S declare that they have no conflict of interest.
Data sharing statement: The datasets generated and/or analyzed during the current study are not publicly available due to legal restrictions under the General Data Protection Regulation and the Bundesdatenschutzgesetz (BDSG). Data may be available from the corresponding author upon reasonable request, provided that access is consistent with the original consent of the participants, approved by the responsible ethics committee, and subject to a formal data use agreement in accordance with applicable data protection regulations.
Corresponding author: Dominik Schulz, Senior Scientist, III. Medizinische Klinik, Universi
Received: March 19, 2026
Revised: May 18, 2026
Accepted: July 3, 2026
Published online: December 7, 2026
Processing time: 210 Days and 12.6 Hours
Revised: May 18, 2026
Accepted: July 3, 2026
Published online: December 7, 2026
Processing time: 210 Days and 12.6 Hours
Core Tip
Core Tip: Accurate measurement of colorectal polyp size is mandatory influencing the choice of resection technique and surveillance intervals. However, conventional visual estimation by endoscopists is prone to significant error rates up to 66%. In this work, we compare three state-of-the art monocular depth estimation artificial intelligence models for automated size measurement. When applied natively to colonoscopy, they show poor accuracy in size prediction. By domain specific training using virtual colonoscopy images we could significantly improve size prediction. The fine-tuned model achieved a mean absolute percentage error of 33.9%-37.0%. This performance was comparable to trainee endoscopists but remained inferior to experts.