Accelerating endoscopic diagnosis by videomics: The synergy of artificial intelligence and clinical endoscopy

Abstract

Videomics, an emerging interdisciplinary field, harnesses the power of artificial intelligence (AI) and machine learning (ML) for the analysis of videoendoscopic frames to improve diagnostic accuracy, therapeutic management, and patient follow-up in medical practice. This article reviews recent advancements and challenges in the application of AI and ML techniques, such as supervised learning, self-supervised learning, and few-shot learning, in videomics for otolaryngology-head-and-neck surgery. We discuss key concepts and tasks in videomics, including quality assessment of endoscopic images, classification of pathologic and nonpathologic frames, detection of lesions within frames, segmentation of pathologic lesions, and in-depth characterization of neoplastic lesions. Furthermore, the potential applications of videomics in surgical training, intraoperative decision-making, and workflow efficiency are highlighted. Challenges faced by researchers in this field, primarily the scarcity of annotated datasets and the need for standardized evaluation methods and datasets, are examined. The article concludes by emphasizing the importance of collaboration among the research community and sustained efforts in refining technology to ensure the successful integration of videomics into clinical practice. The ongoing advancements in videomics hold significant potential in revolutionizing medical diagnostics and treatment, ultimately leading to improved patient outcomes.