Your voice can say a lot about you. Police investigators are trained to pick up on subtle clues in a suspect’s speech patterns in order to ascertain whether or not they are lying, but all of us can do this to some extent. Humans are highly sensitive to non-verbal clues in the speech of others, allowing us to sense the moods and intentions of the people we interact with on a deeply subconscious level, even if they are speaking a foreign language. Studies show that the ability to recognise familiar voices is a skill we acquire in early infancy.
While we usually think of the information conveyed in our voices as something that can catch us out or give us away, what if it could be turned to our advantage? What if our voices are providing information that even we are not yet aware of, such as the presence of a disease? One of the most exciting areas of medical research currently underway is the investigation of vocal biomarkers. Scientists believe that analysis of our speech can provide early warning signs of both psychological and physical disease, allowing these conditions to be diagnosed and treated as quickly as possible. From spotting PTSD in war veterans to identifying the presence of heart disease or Parkinson’s, the possible applications of this technology are truly amazing. In a few short years it may be possible for patients to self-diagnose serious conditions in a matter of seconds simply by recording a sample of their own speech. Let’s take a look at how this technology works, and whether it can be scaled up for worldwide use.
How does it work?
Many former soldiers suffer from psychological conditions such as PTSD and depression. These frequently go undiagnosed as patients who have spent time in the army are often reluctant to talk about their mental health. In 2015 Dr Charles Marmar, a psychologist from NYU who specialises in PTSD, embarked on a five year study into the use of vocal biomarkers as a diagnostic tool for psychological conditions. Marmar is collecting voice samples from both veterans and control group subjects and, using machine learning, has been able to identify 40,000 separate measureable vocal features. Of these, 30 appear to be accurate indicators of the presence of PTSD. In an early experiment, these vocal biomarkers were found to be 77% accurate in distinguishing between PTSD sufferers and healthy volunteers.
While the analysis of vocal biomarkers can be used as a sort of a lie detector test to find psychological conditions that a patient doesn’t want to admit to having, it can also flag up physical conditions of which the patient is completely unaware. The Mayo Clinic has teamed up with Israeli company Beyond Verbal in order to study the use of vocal biomarkers in the diagnosis of heart disease. In an initial study, 150 people were asked to submit three short recordings of their own voice using an app developed by Beyond Verbal. The company then analysed the clips using machine learning and were able to isolate 13 separate vocal traits linked to coronary artery disease. Blockages in the arteries can affect speech in a way that is imperceptible to the human ear, but can be easily detected by a computer. Amir Lerman, a cardiologist from the Mayo Clinic explains:
“What we found out is that specific segments of the voice can be predictive of the amount or degree of the blockages found by the angiography.”
Most excitingly, the fact that the tests were conducted using a simple smartphone app means that diagnosis of heart disease based on vocal biomarkers is highly scalable. In the near future, patients could potentially self-diagnose using the app, adding their voice recording to a growing bank of samples in the process. This library of voices could be used to increase the accuracy of diagnoses over time. The next step for the Mayo Clinic is to conduct similar tests in China to see if the same vocal biomarkers apply in tonal languages as well as non-tonal ones. This video explains the project in more detail.
Heart disease isn’t the only condition linked to vocal biomarkers. Parkinson’s disease is highly debilitating and, unfortunately, is impossible to spot using a blood test. However, research shows that it affects the patient’s voice even in its very early stages. The Parkinson’s Voice Initiative aims to record 10,000 voices in order to create a vast reference library of vocal characteristics that can be used to diagnose Parkinson’s accurately and cheaply. Best of all, it’s easy to help out. Anyone with an Android smartphone can participate and be part of medical history!
A helping hand for innovators
Vocal biomarkers could prove to be an invaluable diagnostic tool, completely changing the way we look for the early signs of disease. However, there is a lot of work to be done before this technology can be rolled out for general use. If your company is working on ways to improve medical diagnostics, or any other area of medicine for that matter, you should consider applying for an Innovate UK Grant. The organisation is currently running the first round of its 2018 Biomedical Catalyst competition. UK SMEs can apply for between £250,000 and £4 million in early stage funding for projects that aim to improve healthcare outcomes using new technology. The competition is open to projects covering all aspects of medicine, including but not limited to:
- disease prevention and proactive management of health and chronic conditions
- earlier and better detection and diagnosis of disease, leading to better patient outcomes
- tailored treatments that either change the underlying disease or offer potential cure
Because Innovate UK Grants are awarded competitively, your application will need to be as eye-catching as possible. You’ll have to convince the judges that your project is more deserving of funding than other rival projects. That’s where we can help. Our experts have years of experience with Innovate UK grants and can guide you through the entire application process.
Get in touch today and let’s get started.