Expected Values of Artificial Intelligence in Colon Capsule Endoscopy: Artificial Intelligence in Capsule Endoscopy...

The purpose of this position statement is to propose expected performance values for artificial intelligence (AI) applications in colon capsule endoscopy (CCE), with emphasis on detection, characterization, localization, reporting, and downstream management of colonic neoplasia. The task force was created within the Artificial Intelligence in Capsule Endoscopy (AICE) consortium project work and drew on members of the international CApsule endoscopy REsearch (iCARE) group for development, review, consensus voting, and final approval. These values are intended as consensus-based benchmarks for research, validation, and future clinical translation, while explicitly acknowledging the limited availability of robust CCE-specific clinical evidence. The task force included CCE experts, technical experts, patient representatives, and an ethicist, and defined expected values using existing endoscopy quality frameworks, available capsule endoscopy literature, relevant proxy evidence from colonoscopy, and structured Delphi consensus. A supermajority agreement of more than 80% across up to three Delphi voting rounds was required for consensus. The agreed statements address candidate selection, assessment of mucosal visualization and procedure completeness, landmark and segment recognition, polyp matching, lesion size estimation, reading-time reduction, detection of colorectal neoplasia, false-positive burden, risk stratification for subsequent procedures, prediction of advanced pathology, urgency categorization, and automated report generation. The proposed thresholds should not be interpreted as definitive requirements for immediate clinical adoption. Instead, they identify minimum expected value and aspirational value that future systems should be tested against in prospective, externally validated, full-video CCE studies.