Expected values of artificial intelligence in colon capsule endoscopy: I-CARE group and AICE consortium Position Statement

Aims The objective of this Position Statement is to outline the primary outcomes of artificial intelligence (AI) tasks within the context of established and emerging performance measures, projecting minimum and desirable values expected during the implementation of AI in colon capsule endoscopy (CCE) practice, focusing on the detection and characterization of colonic neoplasia. The assumption is that AI implementation can standardize quality metrics in CCE with an emphasis on clinical rather than technical validation. Methods The AI-Supported Image Analysis in Large Bowel Camera Capsule Endoscopy (AICE) consortium and the CApsule Endoscopy REsearch (i-CARE) group have established a Task Force (TF) of CCE experts, technical information experts, patients representatives and an ethicist to define the expected values of AI performance measures in CCE, particularly focusing on GI neoplasia, within the framework of previously defined proxy performance measures, primarily conventional colonoscopy. To align AI tasks with performance measures and define their expected values, the existing methodology [ 1 ] was adopted, including task definition, integration of AI tasks with performance measures, the reference standard, and finally the expected value assessment defined for each outcome based on the clinical relevance of the performance measure related to the AI task and the expected benefit/harm of AI implementation. A super agreement of>80% on a maximum of three Delphi voting rounds using a 5-point scale ranging from "strongly disagree" to "strongly agree" was required for consensus on each statement. This project is funded by the European Union. Views and opinions expressed are, however, those of the authors only and do not necessarily reflect those of the European Union or the European Commission. Neither the European Union nor the European Commission can be held responsible for them. Results The TF identified eight domains, namely pre-procedure, completeness of procedure, identification of pathology, management of pathology, complications, patient experience, post-procedure, and ethics. The TF defined 21 AI tasks (shown in the [ Table 1 ]) with their performance measures, reference standards, and expected values. TThe Delphi method was employed to solicit votes on proposed statements regarding the acceptance of AI-assisted CCE in clinical practice. A super agreement was reached in the first round for 17 statements, while the remaining four were amended and approved in the second round. Table 1 Domain Pre-procedure Completeness of procedure Identification of pathology Management of pathology Complications Patient experience Post-procedure Ethics AI Tasks 1. Prediction of failed CCE or need for subsequent examination 2. Adequate colon cleansing 3. Complete mucosa visualization 4. Identification of colon landmarks-segments 5. Correct assignment of pathology to colon segments 6. Polyp localization & CCE path reconstruction 7. Visualization of the small bowel 8. Reading time 9. Polyp matching 10. Polyp size measurement 11. Sensitivity & specificity for: CRC, and polyps 12. False positive rate 13. Non-malignant pathology detection 14. PIVI/SODA criteria for not recommending colonoscopy 15. Prediction of submucosal invasion 16. CCE retention 17. Colonic perforation 18. AI recognition or recording of patients' experience before, during, and after CCE 19. AI-generated report 20. AI-provision of alarm and interface with the clinician and patient. 21. AI-related ethics issues Conclusions A Task Force of clinical and AI technology experts, patients' representatives, decision-making experts, and ethicists described CCE AI tasks and their performance measures and defined statements for acceptance of AI-assisted CCE in clinical practice, focusing on the detection of colorectal neoplasia. Publication History Article published online: 05 June 2026 © 2026. European Society of Gastrointestinal Endoscopy. All rights reserved. Georg Thieme Verlag KG Oswald-Hesse-Straße 50, 70469 Stuttgart, Germany