(0105) Establishing Research Priorities for Artificial Inteilligence Implementations in Upper Aerodigestive Tract Endoscopy: A Modified Delphi Study

Introduction: Artificial intelligence (AI) applications in videoendoscopy for the upper aerodigestive tract (UADT), including segmentation, diagnosis, and treatment planning, are emerging in research but are still not part of clinical practice. This paper introduces an international expert consensus statement (ECS) delineating the priorities for implementation research in this field.

Methods: A modified Delphi method was followed for establishing research priorities. Clinical experts, including otolaryngologists of various subspecialties and speech-language pathologists, who had at least one peer-reviewed publication in AI research utilizing UADT endoscopic images, participated in 3 rounds of online surveys and virtual discussions. The group generated questions on computer vision research utilizing UADT, focusing mainly on the clinical implications. In subsequent rounds, the list was scored to identify the top 10 research priorities and compile statements that reached the standardized consensus definition.

Results: After 3 iterations of Delphi surveys and the removal of duplicate statements, the top 10 research priorities were identified. High-yield topics and questions were grouped into the following themes: data access and annotation, clinical integration and settings, and performance and representativity measures.

Conclusions: This is the first international expert consensus on implementation priority setting in AI research on UADT endoscopy. These statements can be used to guide future research and catalyze the implementation of AI in UADT endoscopy.