Crowdsourced voice recordings: towards a disruptive technology for Parkinson's disease


Many of us know someone who has Parkinson's — it affects more than 6 million worldwide — and would love to play an active role in research into treatments. That's my situation: my friend Jan Stripling (left), who used to be a virtuoso ballet dancer, was recently diagnosed with Parkinson's. For someone whose entire existence revolved around the joy of high precision, coordinated movement that took a sizable fraction of his lifetime to perfect, this incurable disease, that replaces voluntary movement with tremor, rigidity and weakness, is cruel.

As a scientist, I'm fortunate to perhaps be in a position to help. But here I want to argue that anyone can help! Although there are no known biomarkers, research by me (and others) shows that with voice recordings alone, we can quantify symptoms of the disease (e.g. reproduce symptom measurement on the standard clinical scale with around 2% error), and also detect when someone has the disease (with around 99% accuracy). To adapt this technology to the standard phone network will require 10,000 voice recordings from healthy controls and people with Parkinson's.

Last month, we launched the Parkinson's Voice Initiative (PVI) to crowdsource these 10,000 calls. The response has been truly amazing — nearly 5,000 people have contributed — we are overwhelmed! It seems: if you provide a simple, quick (around 3 minute), anonymous, and painless way to contribute to scientific research, people will happily participate. And that gives me tremendous hope for the future of medical science.

If the PVI reaches its goals, the technology we can build with those recordings could mean several big advances for Parkinson's, including reducing the need to visit the clinic for symptom checkups. But perhaps more revolutionary is the impact this could have on the search for new treatments. One of the major stumbling blocks on the road to new drugs is the cost and inconvenience of recruiting large numbers of subjects to get sufficient statistical power in the results. If the only action needed to follow the progress of subjects is a low-cost, 3-minute phone call, then we can run clinical trials on a truly massive scale (think: millions of subjects for overwhelmingly definitive scientific conclusions).

And finally, there's tantalizing, but so far weak, evidence that voice disturbances may be an early disease symptom, present long before the tell-tale limb tremors of Parkinson's. If this is true, voice could become an "acoustic biomarker" of the disease. Why is this so important? By the time the tremors of Parkinson's lead to clinical diagnosis, the neuronal damage is usually beyond repair. So, the phone could become a radically cheap way to screen for the early symptoms, in the general population. We could then work out who might best participate in trials for potential cures, rather than trials for drugs that just suppress the symptoms temporarily. While this "voice as early biomarker" hypothesis is a long shot, we now have the means to test it and think it's too great an opportunity not to pursue.

Parkinson's Voice Initiative