Artificial intelligence in otolaryngology: from the laboratory to clinical practice

Abstract

Artificial intelligence (AI) is no longer a futuristic concept but a reality that is redefining medical practice. In otolaryngology, this technological revolution is beginning to take shape, although challenges still remain. What just a decade ago seemed like an aspiration—such as a machine interpreting an otoscopy or predicting a malignant dysphonia—now has concrete results, robust publications, and an increasingly active scientific community.

Recent evidence is compelling. In otology, multiple studies have shown that deep learning algorithms can diagnose otitis media or tympanic membrane perforations with accuracies around 90%. In laryngology, AI models applied to endoscopic image analysis reach sensitivities close to 80% in detecting malignant lesions, while in rhinology they have been used to identify nasal polyps and evaluate sinus anatomy, showing high agreement with expert judgment. Even in surgical education, AI is beginning to play a teaching role, assessing skills in mastoidectomies or endoscopic surgeries with a precision that rivals that of human instructors.

But beyond the numbers, what matters is the direction these advances indicate. AI promises to optimize clinicians’ time, improve diagnostic accuracy, and, above all, offer more personalized and equitable medicine. However, a considerable gap persists between promise and practice. Most studies have been developed in controlled environments, with limited databases and without large-scale clinical validation. This is what some authors refer to as the “AI chasm”: a space between technical feasibility and real-world applicability.

Other challenges add to this: the lack of well-documented multicenter databases, the need for explainable and transparent algorithms, and the urgency of training specialists in digital competencies. Moreover, the ethical and legal dilemmas accompanying this process cannot be overlooked: data protection, responsibility in AI-assisted decision-making, and the risk that algorithmic biases reproduce inequities already present in healthcare.

The question, therefore, is not whether AI will transform otolaryngology, but how it is expected to do so. The answer should not be technological, but ethical and professional. AI must be conceived as a tool that amplifies the physician’s diagnostic capacity, not as a substitute for their clinical judgment. Its role is to accompany the specialist, serve as a “clinical copilot,” and provide efficiency and objectivity without displacing the empathy or intuition that define the medical act.

To advance toward this integration, the scientific community must promote interdisciplinary collaborations that bring together clinicians, engineers, and data researchers. It is imperative to encourage clinical validation studies, develop interoperable platforms, and ensure that AI is built on principles of transparency, equity, and evidence. And of course, digital literacy must be included in the training of future otolaryngologists so that they understand both the potential and the limits of this technology.

AI is not here to replace the otolaryngologist’s expert eye, but to offer a new lens. If innovation can be balanced with ethics, and technology with human sensitivity, the result will be more precise, more predictive, and—paradoxically—more human medicine.