Building Trustworthy Vocal Biomarkers from Mechanisms to Clinical Impact

Voice is an accessible, dynamic biosignal that can carry information about physiology, behavior, cognition, and health. Changes in phonation, articulation, prosody, fluency, language use, cough, and breathing may accompany neurological, mental health, respiratory, cardiovascular, infectious, and other conditions. Together with advances in smartphones, telehealth, wearable sensing, signal processing, and artificial intelligence, this creates new opportunities for low-burden screening, phenotyping, prognosis, longitudinal monitoring, and assessment of treatment response.However, high predictive performance on a single dataset is not sufficient for clinical adoption. Promising vocal biomarkers must be linked to plausible biological or behavioral mechanisms, measured reliably, validated against appropriate clinical reference standards, and shown to generalize across people, languages, devices, recording tasks, and care settings. They must also be interpretable enough to support clinical reasoning, responsibly governed, and designed around the needs of patients, clinicians, and health systems.This Research Topic brings together clinical, technical, and translational perspectives to move vocal biomarkers beyond proof-of-concept prediction toward trustworthy and actionable health tools. We invite contributions from medicine, speech-language pathology, acoustics, linguistics, engineering, data science, public health, human factors, ethics, regulatory science, and industry. Submissions may address any stage of the pathway from mechanism and discovery to analytical validation, clinical evaluation, implementation, and real-world impact. Studies of voice, speech, language, cough, breathing, and related human-generated audio are welcome, either alone or integrated with electronic health records, wearables, imaging, laboratory measures, omics, or other digital biomarkers. Work across diseases, life stages, languages, cultures, care settings, and resource contexts is encouraged. We welcome all article types accepted by the participating Frontiers sections. Relevant themes include, but are not limited to:1. Mechanistic foundations. Biological, physiological, neurological, cardio-respiratory, and behavioral mechanisms underlying health-related changes in voice, speech, language, cough, or breathing.2. Explainability and clinical meaning. Mechanism-informed and explainable models; clinically meaningful acoustic, linguistic, and learned representations; uncertainty estimation, calibration, and transparent communication of model outputs.3. Analytical validation and standardization. Reliability, repeatability, signal quality, task design, sensitivity to recording hardware and software, and standardized collection, preprocessing, feature extraction, and reporting protocols.4. Clinical validation and utility. External, prospective, multicenter, longitudinal, or pragmatic studies using appropriate reference standards and evaluating diagnostic, prognostic, monitoring, or treatment-response utility.5. Generalizability, fairness, and transportability. Performance across languages, accents, demographic groups, comorbidities, devices, channels, environments, and health systems, including bias assessment, mitigation, domain adaptation, and transportability.6. Longitudinal and real-world monitoring. Remote assessment, disease trajectories, relapse or exacerbation detection, symptom severity, treatment response, and personalized care using repeated or continuously collected vocal data.7. Multimodal and emerging AI. Self-supervised, foundation, and generative AI approaches that integrate voice, speech, and language with complementary clinical, physiological, imaging, laboratory, behavioral, or wearable data.8. Data resources and reproducible science. Dataset design, metadata, quality control, benchmarks, interoperable and reproducible pipelines, federated or privacy-preserving learning, and responsible data access and sharing.9. Responsible implementation, regulation, and translation. Human factors, patient and clinician acceptability, privacy, consent, governance, and regulatory pathways and academic-clinical-industry collaboration, health-economic evaluation, and sustainable pathways from research prototypes to deployable products and services.10. Replication and boundary conditions. Replication studies, cross-dataset benchmarking, reporting standards, and analyses of failure modes that clarify when, where, and for whom vocal biomarkers are clinically informative.By emphasizing clinical meaning alongside technical performance, this collection aims to define the evidence, methods, and partnerships required for vocal biomarkers to become trustworthy, equitable, and clinically useful. Topic editors: Jas Sara, Daria Hemmerling, Marcus Hott Source: https://www.frontiersin.org/research-topics/85344/building-trustworthy-vocal-biomarkers-from-mechanisms-to-clinical-impact