By David Ryan Polgar, Contributor
Are you okay? You don’t sound well.
The quiver in a person’s voice and changes in their intonation are telltale signs that something is amiss. But now artificial intelligence (AI) is going a step further and offering a warning sign of a potentially underlying disease.
Vocalis Health, a software as a service (SaaS) startup based in Newton, Massachusetts, is leveraging AI and machine learning to analyze a person’s breathing and vocal patterns. According to the company, the subtleties in the voice can give insight into whether someone is more likely to have COVID-19, or if they’re at higher risk when it comes to pulmonary hypertension and congestive heart failure. The company is currently testing a mobile app with patients and physicians, where a patient records their voice everyday and the physician can monitor fluctuations in shortness of breath.
Vocalis Health relies on vocal biomarkers, a technique to analyze voice patterns and other “markers” that may be correlated with underlying health concerns. A biomarker, according to the National Center for Biotechnology Information, is an objective medical sign observably outside the patient that can be measured accurately. Shortness of breath is a common vocal biomaker related to a patient’s medical state.
This is VocalisTrack’s first chronic-care product. Future patients will record their voice into the VocalisTrack app each day, and their physicians would be provided a dashboard that would analyze the fluctuations in any shortness of breath and pinpoint areas of concern. VocalisTrack is currently in its beta phase awaiting regulatory clearance in the European Union (E.U.). The company plans to continue to conduct clinical pilots in the E.U. and the U.S., with an eye toward limited market commercial releases.
In the future, the company plans to expand into more predictive analytics, finding possible correlations between vocal biomarkers and neurological disorders, sleep apnea, and depression.
The Voice as a Gateway to Understanding Our Health
While a patient is often able to detect symptoms of an underlying health concern or describe how they subjectively feel, they may have biological abnormalities or fluctuations in their voice that they wouldn’t necessarily notice. The human ear has a hearing range between 20 Hertz (Hz) and 20,000 Hz, and is unable to detect infrasounds—below 20Hz—and ultrasounds—above 20,000 Hz—that fall outside of audible hearing. Subtle changes in shortness of breath may go unnoticed by an individual, but could be flagged by a system that is analyzing speech. COVID-19, which has become a major focus for Vocalis Health, affects the respiratory system, making vocal biomarkers a potential source of detection.
“Voice had not been used systematically or in a standardized way across the healthcare system,” says Tal Wenderow, president and CEO of Vocalis Health. He points out that the healthcare system already utilizes biomarkers, such as a patient’s blood pressure, ability to smell, and changing skin color.
What We Can Learn Through Our Voice
“What is going on in our vocal machine is constantly giving up a lot of information,” says vocal expert Marshall Davis Jones, founder of MindBodySpeak. “The entire vocal system is a feedback loop in ways we wouldn’t think.” According to Jones, research from 2019 supports voice as deeply connected to the rest of the body. It can be a gateway to greater understanding about overall well-being, says Jones. “We are looking at the multi-modal transference of data even across the senses. Now we are recognizing that all of these systems are working synergistically—that information combined and aggregated could give a much more sophisticated readout of where someone is.”
“What is going on in our vocal machine is constantly giving up a lot of information. The entire vocal system is a feedback loop in ways we wouldn’t think.”
—Marshall Davis Jones, founder, MindBodySpeak
Currently, vocal biomarkers would not be used to diagnose a disease, but rather act as a first line of defense. Given the mainstream adoption of smartphones with recording technology and a growing comfort of speaking to devices, Vocalis Health would equip patients with the ability to record their voice into a dashboard that is tracked by a healthcare professional.
Wenderow sees shortness of breath as the lowest hanging fruit for vocal biomarker detection software like Vocalis Health to showcase its potential. He sees Vocalis’ first product that examines shortness of breath as a way to make physicians comfortable with the company’s technology. Since physicians have experience correlating a patient’s shortness of breath with an underlying health concern, it is easier for them to understand—and therefore trust—the information that Vocalis is presenting.
How Using an App Could Help Detect an Underlying Health Problem
While the process of using algorithms as a predictive measure for health may seem futuristic, Wenderow points out that some companies train their algorithms based on image and medical diagnosing and others based on blood count and medical diagnosis. For example, researchers from the Institute of Advanced Study in Science and Technology in India recently announced that they have developed software to detect cervical cancer through the analysis of Pap smear images. Similarly, Vocalis Health is training its algorithm on voice and the medical diagnosis, and looking for statistically relevant correlations. This information could be gathered and analyzed through an app, such as the company’s VocalisTrack.
It’s been known that our voice unlocks our emotional state. If Wenderow and Vocalis Health are successful, using our voice may unlock our medical state, as well. Instead of just leveraging our voice to turn on the lights or find out the weather, it may tell you that it’s time to go see a doctor.