Research / Voice as a clinical signal / [7]
Voice biomarkers need clinical-grade validation: extending the V3 framework
Dataset-Driven Voice Biomarker Validation: Advancing Clinical Standards with the Enhanced V3 Framework. XoveTIC 2024 (7th XoveTIC Congress, A Coruña), 2024.
Voice is a promising biomarker for disease detection, but it lacks rigorous clinical validation methods. The paper refines the V3 framework for voice and proposes an open-source toolkit to evaluate voice health datasets.
The problem
A voice biomarker is a measurable feature of a person’s voice that indicates a health condition. Voice is a promising biomarker for detecting disease, but its use in clinical practice is held back by the lack of rigorous validation methods. Research results do not turn into clinical tools without a framework that says what has to be shown, and how.
What the paper proposes
It builds on the V3 framework and on the authors’ experience running a clinical trial, and adapts the framework to the specific problems of voice biomarkers. The focus is the data life cycle in a trial: how recordings are collected, handled and checked.
Alongside the refined framework, the paper proposes an open-source toolkit for evaluating voice-based health datasets.
Why it matters
A voice model that scores well on one dataset says little about clinical use. A shared validation procedure is what lets results from different studies be compared and trusted by clinicians.
Cite
Ramírez J.M., Docío-Fernández L., García-Mateo C. Dataset-Driven Voice Biomarker Validation: Advancing Clinical Standards with the Enhanced V3 Framework. XoveTIC 2024 (7th XoveTIC Congress, A Coruña), 2024.
PDF distributed under CC BY 3.0 ES; copy of the published version.