Voice disorders classification using machine learning: a scoping review.

Area of Science:

• Medical Informatics
• Biomedical Engineering
• Artificial Intelligence in Healthcare

Background:

• Machine learning (ML) shows promise for voice disorder classification.
• Clinical application of ML for multi-class voice disorders is limited.
• Standardization is needed for reliable ML models in voice pathology.

Purpose of the Study:

• Identify key barriers to clinical application of ML in multi-class voice disorder classification.
• Analyze patterns in ML techniques for voice disorder classification.
• Highlight inconsistencies hindering the translation of ML research to clinical practice.

Main Methods:

• Conducted a comprehensive scoping review of research from 2013 to May 2025.
• Included articles applying ML for multi-class voice disorder classification.
• Extracted data on classification classes, databases, input attributes, vocal tasks, labels, and ML techniques.

Main Results:

• Screened 10,401 articles; 80 used ML for multi-class classification.
• Observed significant variation in databases, diagnostic labels, input data, and ML techniques.
• Inconsistencies prevent robust comparisons and identification of state-of-the-art solutions.

Conclusions:

• Variations in diagnostic labels, data, and testing methods limit ML model comparability and generalization.
• Lack of consensus in the automated classification pipeline impedes clinical application.
• Addressing these barriers is essential to realize voice as a biomarker for diseases.