Artificial Intelligence in Otolaryngology: Current Applications, Limitations, and Future Perspectives

The review systematically organizes the clinical application landscape of AI across multiple otolaryngology subspecialties, covering diagnostic imaging, endoscopic assessment, audiology, rhinology, laryngology, vestibular medicine, surgical simulation, head and neck oncology, radiotherapy planning, and predictive analytics. Compared with studies focused on a single subspecialty or algorithm, this review integrates scattered applications into a cross-subspecialty clinical map, helping readers grasp the field's overall progress. Evidence comes from a structured literature search across three databases, prioritizing English peer-reviewed studies with clinically relevant diagnostic, prognostic, surgical, educational, or workflow applications, representing a narrative-review level of synthesis.

The review indicates that deep learning algorithms show promise in sinonasal disease detection, automated image segmentation, lymph node metastasis prediction, thyroid nodule classification, and prognostic modeling for head and neck cancer patients. These directions previously appeared mainly as individual studies; the review consolidates them into representative categories with clinical translation potential. The text uses phrasing such as 'emerging evidence suggests,' indicating these conclusions rest on preliminary or emerging evidence rather than established clinical consensus.

The review highlights that multimodal AI systems integrating radiologic, pathologic, molecular, and clinical data may further improve diagnostic precision and personalized treatment planning, and that generative large language models show emerging applications in medical education, image interpretation, differential diagnosis, and clinical decision support. By including multimodal fusion and generative models, the review extends the discussion beyond the predominantly discriminative models covered in earlier accounts. This is a summary description of emerging directions; the text provides no specific performance metrics or comparative data.

The review consolidates key challenges limiting broad clinical implementation, including limited external validation, retrospective study design, dataset heterogeneity, algorithmic bias, lack of transparency, patient privacy concerns, medico-legal uncertainty, and automation bias risks. By listing methodological and translational barriers together, it offers a problem checklist for future study design and governance frameworks. This is a synthetic judgment based on the overall characteristics of the included literature rather than a single empirical result.

This article is positioned to provide clinicians, researchers, and relevant administrators in otolaryngology-head and neck surgery with an application overview and problem checklist, suitable for understanding the current state of AI applications in diagnosis, prognosis, surgery, education, and workflow in this field. Its conclusions are framed around AI as an 'adjunctive tool,' and the authors explicitly state that broad clinical adoption is not yet appropriate until prospective multicenter studies, standardized validation frameworks, and ethical oversight are in place.

Readers should note that the evidence strength varies considerably across the application directions described, and is presented with phrasing such as 'emerging evidence' and 'may,' so specific performance and external validation require consulting the original studies; applications of multimodal systems and generative large language models remain early-stage, with clinical value yet to be further verified; moreover, the available text is a condensed version of the review body without figures or reference details, so no judgment can be made about specific study counts, sample sizes, or algorithm metrics.