Using voice to diagnose disease: collaboration uses machine learning to aid diagnostics
A collaboration of 10 universities, including the University of Toronto, has recently been awarded funding under the National Institutes of Health’s Bridge2AI program, designed to use AI to tackle complex biomedical challenges. Led by the University of South Florida and Weill Cornell, two faculty appointed in Laboratory Medicine and Pathobiology in the Temerty Faculty of Medicine will play roles in the project.
Changes in your voice can be indicative of different types of diseases like Alzheimer's or Autism. Called “Voice as a Biomarker of Health”, the funded project aims to build an ethically sourced database of diverse human voices. Using this data, machine learning models will be trained to spot diseases by detecting changes in the human voice, which could provide doctors with a low-cost diagnostic tool to be used alongside other clinical methods.
The group of medical, voice, AI, engineering, and ethics experts will study disease categories such as:
- 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)
Dr. Jordan Lerner-Ellis is an Associate Professor in LMP and co-Head and co-Director of the Advanced Molecular Diagnostics Laboratory at Mount Sinai Hospital. The laboratory offers clinical genetic testing services for the province of Ontario and his research team has a long-standing interest in understanding genomic data and the human genome.
Lerner-Ellis is the Genomic Cohort Lead and his role in the project will be to work with clinics in Toronto and enroll patients to collect samples for the purpose of genome sequencing.
“Our aim is to build a publicly available database that researchers from around the world can use to study the relationship between genetics and disease and voice disorders; this approach could eventually lead to the development of valuable diagnostic tools. We will look at the relationship between genetic variation and the voice. Where there is a known causal molecular biomarker, such as for Alzheimer's, we can return this information to clinicians and research participants,” explains Lerner-Ellis.
His team will play a critical role in interacting with participants. “We do a lot of genome sequencing and returning these results to participants can provide important information such as facilitating a diagnosis and helping to inform prognosis and patient management. Our team will be directly engaged in patient enrollment and consent, sample collection, genome sequencing and pre- and post-test genetic counselling,” he says.
Dr. Frank Rudzicz, is an Associate Professor in LMP and the Department of Computer Science, and Scientist at the Li Ka Shing Knowledge Institute at Unity Health Toronto. An expert in machine learning in healthcare, especially in natural language processing, speech recognition, and surgical safety, he has been working on speech analysis in Aphasia and Dementia so was keen to get involved in the project.
Leading the Neurological Cohort, he is part of the data acquisition team, focused on behavioral, speech and language aspects.
“When datasets are too small, the generalizability of models trained on those datasets becomes a bit of a problem. With this study, we’ll be able to look at a broad range of diseases and with the data we'll be able to get a much deeper sense of the phenomena.”
“This project is ambitious and very large. We cover everything from natural data collection and analysis to ethical considerations. We also overcome one of the key challenges in this kind of science, which is siloed, private data sets. If data are private, it’s not really possible to reproduce the results so, in the interests of science, having clean, large, public data sets which can be used as a baseline for many different methods is so important. I'm really thrilled about that aspect of this project in particular,” says Rudzicz.
Bridge2AI: Voice as a Biomarker of Health - Building an ethically sourced, bioaccoustic database to understand disease like never before
Award number: OT2 OD032720 Voice as a Biomarker of Health is being co-led by Dr 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 and Gaetane Michaud (USF Health Morsani College of Medicine), Alistair Johnson (Hospital for Sick Children), 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), Kathy Jenkins (Boston Children’s Hospital), Frank Rudzizc (University of Toronto) and Jordan Lerner-Ellis (Sinai Health). AI startup Owkin is deploying its federated learning technology across multiple research institutions to protect the security and privacy of sensitive voice data.
Artificial Intelligence in Healthcare at U of T
To find out more about how the U of T is facilitating medical research through data sharing, learn about the Health Data Nexus
The Temerty Centre for Artificial Intelligence Research and Education in Medicine (T-CAIREM) at the University of Toronto