How Well Will AI Help Recognize Voice Disorders? A State-of-the-art Review of Current Acoustic Assessment Strategies and Future Applications.

Objective: To discuss the current clinical application and usefulness, shortcomings and future directions of traditional and artificial intelligence (AI)-driven acoustic assessment techniques to detect voice dysfunction. Data Sources: Literature review. Conclusion: AI-based acoustic voice analysis techniques have huge potential to improve the early recognition, diagnosis, and tracking of treatment success in patients with voice disorders or diseases affecting voice function. Through smartphones, wearable devices, and server-based solutions, acoustic voice assessment techniques have become widely available and may be extended to workplace and private settings. However, the transformative potential is thwarted by several limitations including a lack of (a) consistent data collection and reporting standards, leading to heterogeneity of current databases and literature; (b) characterization what acoustic analysis techniques including AI can detect or track reliably, and whether the derived outcomes serve as a reliable marker of dysfunction, pathology, or an improvement thereof; (c) clinical validation studies in unselected patients; and (d) ethical and legal controversies. Thus, substantial effort to research, define and establish guidelines for the collection, storage, and processing of acoustic data and valid clinical applications is warranted to design sensible strategies for analysis and use.