A national databank of de-identified voices, combined with artificial intelligence, could lead to diagnosing and treating cancer, depression, autism, Alzheimer’s disease and voice disorders.
Vanderbilt University Medical Center is partnering with 11 institutions on a $14 million NIH-funded project led by the University of South Florida and Weill Cornell Medicine that aims to establish voice as a biomarker used in clinical care.
Called Voice as a Biomarker of Health, the project is one of several recently funded by the NIH Common Fund’s Bridge2AI program, designed to use AI to tackle complex biomedical challenges.
The voice project will build an ethically sourced, de-identified database of diverse human voices.
Machine learning models will use the data to identify diseases from the human voice.
“The drive for this data generation project is the critical need for multi-institutional, multimodal, datasets that are accessible to researchers while still maintaining strict standards for patient privacy,” said Maria Powell, PhD, CCC-SLP, assistant professor of Otolaryngology-Head and Neck Surgery at VUMC.
“Our project in particular focuses on acoustic (or voice) data linked to respiratory data, pulmonary and neurological imaging, quality-of-life measures, and other health-related biomarkers that can help us analyze different types of diseases,” she said.
The research team identified five disease cohort categories where voice changes have been associated with specific diseases including:
- Voice disorders: (laryngeal cancers, vocal fold paralysis, benign laryngeal lesions).
- Neurological and neurodegenerative disorders (Alzheimer’s, Parkinson’s, stroke, ALS).
- Mood and psychiatric disorders (depression, schizophrenia, bipolar disorders).
- Respiratory disorders (pneumonia, COPD, heart failure).
- Pediatric voice and speech disorders (speech and language delays, autism).
Powell, who was named principal investigator for the project’s Plan for Enhancing Diverse Perspectives (PEDP), is leading this project with co-investigator Toufeeq Ahmed, PhD, MS, assistant professor of Biomedical Informatics at Vanderbilt University Medical Center, to promote diversity in team recruitment, patient representation and educational programs.
Ahmed brings expertise and leadership to this project from his efforts leading, as principal investigator, the AIM-AHEAD (Artificial Intelligence/Machine Learning Consortium to Advance Health Equity and Researcher Diversity) program, created to enhance participation and representation of researchers and communities underrepresented in the development of AI/ML models and to address health disparities and inequities.
“It is critical for AI projects to consider inclusion, diversity and equity in participant recruitment, data collection practices, AI algorithm and model development, and researcher diversity to advance scientific innovation and biomedical research through the inclusion of all voices,” Ahmed said.
“Our goal is to make sure that, literally, every voice is represented,” Powell said. “From a data generation standpoint, we will focus on identifying and actively recruiting individuals often underrepresented in health care research, to promote inclusivity within our dataset.
“Our PEDP efforts will also focus on promoting diversity in leadership not only within our team, by developing guidelines for inclusive hiring practices, but also within the field, through developing inclusive educational programs,” she added.
Voice as a Biomarker of Health is being co-led by Yaël Bensoussan, MD, from USF Health Morsani College of Medicine, and Olivier Elemento, PhD, from Weill Cornell Medicine, who are co-principal investigators for the project.
The project also includes lead investigators from 10 other universities in North America: Alexandros Sigaras and Anaïs Rameau (Weill Cornell Medicine); Maria Powell (Vanderbilt University Medical Center); Ruth Bahr (University of South Florida College of Behavioral and Community Sciences); Alistair Johnson (SickKids); Philip Payne (Washington University in St. Louis); David Dorr (Oregon Health & Science University); Jean-Christophe Belisle-Pipon (Simon Fraser University); Vardit Ravitsky (University of Montreal); Satrajit Ghosh (Massachusetts Institute of Technology (MIT)); Kathy Jenkins (Boston Children’s Hospital); Frank Rudzizc and Jordan Lerner-Ellis (University of Toronto); Gaetane Michaud (USF Health Morsani College of Medicine).
French-American AI biotech startup Owkin is deploying its federated learning technology across multiple research institutions to protect the security and privacy of sensitive voice data.
Federated learning technology is a novel AI framework that allows machine learning models to be trained on data without the data ever leaving its source, allowing cross-center AI research to be conducted while preserving the privacy and security of sensitive voice data.
The first year of the project includes $3.8 million from the NIH, with subsequent NIH funding over the following three years, bringing the overall award to $14 million.
The project is supported by NIH grant number 1OT2OD032720-01.
