Clinical Evidence

Clinical Evidence for Disease Detection

Research confirms that acoustic features are highly sensitive to physiological and neuromuscular changes across a broad spectrum of conditions.

Neurodegenerative Disorders: Studies in the Framingham Heart Study demonstrate that acoustic features like jitter can predict incident dementia in asymptomatic participants with high accuracy. Additionally, research on Parkinson’s Disease shows that objective measures detect voice changes in up to 78% of early-stage patients.
Respiratory and Systemic Health: Voice biomarkers are now being used to predict Type 2 Diabetes and monitor respiratory conditions like asthma, where pathological voices exhibit higher jitter and lower harmonics-to-noise ratios (HNR).
Vocal Fold Pathology: A 2025 study highlights HNR and its variability as promising markers for the early detection of laryngeal cancer.

Information Density: Acoustic vs. NLP Features

The “5-10x” information density advantage is rooted in the physiological nature of acoustic data compared to the cognitive nature of linguistic (NLP) data.

Multi-Dimensional Analysis: Acoustic analysis extracts hundreds of dimensions—including fundamental frequency (F0), formants, and MFCCs—which capture the "how" of speech (neuromuscular coordination).
Sample Efficiency: Unlike NLP, which often requires longer conversations to analyze lexical diversity or semantic coherence, acoustic biomarkers can be derived from short, standardized speech tasks or even singing.
Superior Accuracy: Modern temporal convolution neural networks using acoustic data have achieved classification accuracies as high as 91.6% across multiple disease classes.