A Comprehensive Review of Diagnostic Approaches to Vocal Fold Paralysis Using Artificial Intelligence

Vocal fold paralysis, a condition caused by nerve damage, leads to impaired vocal fold movement, impacting speech, breathing, and swallowing. Diagnosing this condition is challenging as conventional methods are invasive. Even after requiring specialized equipment, the diagnosis often fails to confidently distinguish it from similar disorders. This results in treatment delay, affecting the patient's quality of life. Artificial intelligence can be trained on datasets to identify patterns in data that humans can miss, especially in cases where the disease incidence is not common. This narrative review consolidates recent research on the application of artificial intelligence on vocal fold paralysis diagnosis in the last five years. All research papers that have aimed to diagnose unilateral or bilateral vocal fold paralysis using artificial intelligence have been reviewed in this work. Datasets, performance, and challenges have been showcased, along with research gaps and areas for improvement. Artificial intelligence models have demonstrated significant diagnostic potential. Artificial intelligence applied to acoustic analysis successfully identified subtle voice changes linked to impaired vocal fold function with high accuracy. Application to imaging-based approaches offered reliable and detailed motion assessments of vocal folds. Computational approaches provide promising supplements if not alternatives to traditional diagnostic tools, enabling earlier identification and personalized treatment. Dataset diversity, model bias, and reliance on high-quality data are consistent challenges. Future research should focus on expanding the pool of available public datasets, refining algorithms, and ensuring usability in clinical settings to maximize the impact of these technologies on patient outcomes.