30-Second Call Would Diagnose Parkinson's
Researchers are working on a system that would help doctors diagnose Parkinson's disease with a 30-second phone call. Callers would speak briefly to the computerized system, which would then automatically analyze callers' voices, searching for telltale tremors.
"We're not intending this to be a replacement for clinical experts, rather, it can very cheaply help identify people who might be at high risk of having the disease," Max Little, a mathematician who is leading the research effort, told the BBC. For people who have already been diagnosed with Parkinson's, the automatic system would help their doctors easily and cheaply track their symptoms in between clinic visits, Little added.
Now people in seven countries around the world can contribute to Little's research by calling into their country's number, listed on Little's research website, and taking a three-minute speaking test designed to create an enormous database of voices of people with and without Parkinson's.
Parkinson's disease is a neurological disease that causes tremors in the limbs and face and stiff, slow, unstable movement. People with Parkinson's get progressively worse with time. Some patients may eventually have trouble walking and talking. There's no cure, though there are drugs to deal with the symptoms, and there's no blood test to diagnose the disease. For now, diagnostic tests requires a 20-minute visit with a neurologist.
In 2003, Little met Intel researchers who were studying Parkinson's. An Intel founder, Andy Grove, has the disease and has donated millions to research. When Little met them, the Intel researchers were using accelerometers in devices to detect the changes in Parkinson's patients' movement, but they had also recorded about 50 Parkinson's patients speaking once a week over six months. "They had an enormous amount of data," Little told the BBC.
Little went to work analyzing the Intel data with his algorithms and found that they were 86 percent accurate at identifying Parkinson's patients.
His computer program works by "learning" the patterns in Parkinson's voices from a set of voices already pre-sorted for it. It then diagnoses new voices using the patterns it learned during training. It's tricky work, as people's voices are also altered by smoking, throat surgery and common colds. Little hopes a large database will improve his program. He aims to gather 10,000 voices.
Little, a TED fellow, asked for volunteer call-ins today (June 25) at the TEDGlobal conference in Edinburgh, Scotland. Here at InnovationNewsDaily, we called in and felt a little silly taking the test in the office, which entails saying, "Ahhhhhh" for as long as possible. It is definitely safe for work, however, and it's just as quick and easy as advertised.