How voice biomarkers have been used in heart failure management: a scoping review

Voice biomarkers have been increasingly applied in disease diagnosis and monitoring because of their non-invasive, low-cost, and easily accessible features. This scoping review aimed to summarize existing evidence on voice biomarkers used in heart failure management and to describe their collection, processing, and analysis methods. Four databases (PubMed, CINAHL, Scopus, and IEEE Xplore) were searched from inception to January 10, 2024. Two reviewers independently screened studies and extracted data using a predesigned form. The review followed the Joanna Briggs Institute framework and the PRISMA-ScR guideline. Nineteen studies published between 2010 and 2023 were included. Voice biomarkers were applied for diagnosis or differential diagnosis (9/19) and for assessing or monitoring disease progression (10/19). Voice/speech (11/19) were the most common biomarkers, typically collected via microphones (6/19), phones (3/19), or applications (3/19). Common processing methods included segmentation (9/19), feature extraction (9/19), denoising (8/19), and resampling (7/19). Most studies (12/19) developed new algorithms for data analysis. Model performance was mainly evaluated by accuracy (10/19), sensitivity (7/19), and specificity (6/19). This review highlights the potential of voice/speech biomarkers in heart failure management. Their non-invasive and accessible nature supports their integration into clinical monitoring. However, standardized procedures for data collection, processing, and analysis should be established through multidisciplinary collaboration to enable reliable clinical application.