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

Area of Science:

• Medical Diagnostics
• Artificial Intelligence in Medicine
• Otolaryngology

Background:

• Vocal fold paralysis, resulting from nerve damage, impairs vocal fold function, affecting speech, breathing, and swallowing.
• Conventional diagnostic methods for vocal fold paralysis are often invasive, challenging, and may not reliably differentiate it from similar disorders, leading to treatment delays.
• Artificial intelligence (AI) offers potential for identifying complex patterns in data that may be missed by human analysis, particularly in less common conditions.

Purpose of the Study:

• To review recent research (last five years) on the application of artificial intelligence in diagnosing vocal fold paralysis.
• To consolidate findings on AI's diagnostic performance, datasets used, and challenges encountered in vocal fold paralysis research.
• To identify research gaps and suggest areas for improvement in AI-driven vocal fold paralysis diagnosis.

Main Methods:

• A narrative review of research papers focusing on AI applications for diagnosing unilateral or bilateral vocal fold paralysis.
• Analysis of studies utilizing AI for acoustic analysis of voice changes related to vocal fold dysfunction.
• Evaluation of research employing AI with imaging-based approaches for vocal fold motion assessment.

Main Results:

• AI models demonstrate significant potential in diagnosing vocal fold paralysis.
• AI applied to acoustic analysis achieved high accuracy in identifying subtle voice changes indicative of impaired vocal fold function.
• AI integrated with imaging techniques provided reliable and detailed assessments of vocal fold motion.

Conclusions:

• Computational approaches, including AI, show promise as supplements or alternatives to traditional diagnostic tools for vocal fold paralysis.
• AI enables earlier identification and personalized treatment strategies for patients with vocal fold paralysis.
• Key challenges include dataset diversity, model bias, and data quality; future research should address these for clinical integration.