Introduction: Artificial intelligence (AI) is increasingly used to analyze acoustic voice data for identifying, classifying, and monitoring voice disorders. Advances in machine learning and signal processing have opened new possibilities for objective voice assessment. However, existing research is fragmented, and there is no systematic synthesis of how AI models have been applied to acoustic parameters in clinical voice evaluation.
Problem Statement: Despite rapid growth in AI applications, the relationship between acoustic features and specific clinical voice outcomes remains unclear. Previous work has focused on single pathologies or isolated algorithms, leaving a gap in understanding how diverse AI methods and acoustic parameters are used collectively to inform diagnosis and monitoring of voice disorders.
Rationale: This project expands on earlier reviews by systematically mapping studies that use AI or machine learning for acoustic analysis of pathological voices. The review aims to identify trends, methodological gaps, and future directions to enhance the rigor and clinical relevance of AI-based acoustic assessment.
Methods: Following PRISMA guidelines, a systematic review is underway in collaboration with a research librarian. Databases include PubMed, Embase, Web of Science, CINAHL, Medline, IEEE Xplore, and Cochrane. Studies are eligible if they apply AI or machine learning to acoustic features (e.g., jitter, shimmer, cepstral peak prominence, MFCC, F0) for detecting, classifying, or evaluating voice disorders. Screening and data extraction are conducted independently by reviewers, with documentation of AI methods, datasets, and performance metrics.
Preliminary Findings: Early screening shows that most studies focus on neurological and structural voice disorders, particularly Parkinson’s disease, with limited research on functional or benign conditions. There is wide variability in acoustic features, dataset sizes, and model validation procedures.
Significance: This ongoing systematic review will provide the first structured synthesis of AI applications in acoustic voice analysis. The findings will help define best practices, promote methodological transparency, and guide future development of standardized AI-driven voice assessment tools for clinical use.