The ‘Voice As a Biomarker of Health’ project will support the training of AI software that can analyse patients voices to diagnose and study illnesses such as cancer, Parkinson’s, and depression.

The US National Institutes of Health (NIH) plans to spend $ 14 million over four years to create a training database of people’s voices that can be used to train AI and machine learning applications for the detection of various diseases and neurological disorders by speech analysis. (The Register)

The money will go to 12 research institutions led by the University of South Florida to fund the “Voice As a Biomarker of Health” for the development of software to detect the following diseases:

· 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)

· Pediatric voice and speech disorders (speech and language delays, autism)

Scientists are of the view that AI and machine learning algorithms have advanced enough in their analysis of voice, speech and breathing data to be able to detect the presence of otherwise difficult-to-diagnose diseases and to assess physical and mental health.

“Voice is one of the cheapest bio-markers to study,” Yael Bensoussan, the project leader and assistant professor at USF’s Department of Otolaryngology, told The Register, raising the possibility that AI/ML could be a viable and cost-effective means for the study of these health problems.

“Our team chose the five categories of diseases based on existing work in voice AI that has been published over the last 20 years,” Bensoussan added. “Voice is the easiest biomarker to collect, does not cause any physical risk for patients, and can be collected in very low resource settings especially with modern technology.”

The project participants will build a large, diverse voice database from speech data that will be recorded in selected patients in clinical settings in a pilot study.

This database will then be used to train AI algorithms to recognise distinctive features in the voices of patients suffering from specific diseases under investigation.

Related Story: AI Used To Create Anthony Bourdain’s Voice In The “Roadrunner” Documentary

Image of “Aerial view of the Clinical Center (Building 10), NIH Campus, Bethesda, MD” – Flickr