British mathematician Max Little and his team have built a speech algorithm that can diagnose Parkinson’s disease from just a phone call, and they need you to help make it better.
Parkinson’s disease is a neurodegenerative disorder affecting about 6 million people worldwide, with more than 50,000 new cases diagnosed each year in the United States. There is no definitive test for the disorder, so it is typically diagnosed through the process of elimination. In the words of Dr. Gregory House, “‘Idiopathic’, from the Latin, meaning we’re idiots ’cause we can’t figure out what’s causing it.” You can’t test for something if you don’t know what’s causing it.
In early tests, Little’s voice-based Parkinson’s screening algorithm managed a 99% accuracy in detecting Parkinson’s speech markers out of 263 recordings of 43 people. The algorithm calculates a dysphonia — the inability to produce sounds using the vocal organ — and measures the symptom severity using the Unified Parkinson’s Disease Rating Scale. This level of accuracy for such a simple test is remarkable compared to current exhaustive methods of diagnosis. However, the algorithm is still in development, and the team intends to build a database of voice recordings from 10,000 people all over the world to fine tune it.
This is machine learning. We are collecting a large amount of data when we know if someone has the disease or not and we train the database to learn how to separate out the true symptoms of the disease from other factors.
The algorithm is able to distinguish other forms of dysphonia caused by a cold or by smoking, but it needs more data to achieve greater accuracy.
Having an accurate diagnostic tool that can be performed at home at very low cost will save patients and researchers a lot of time and money — patients would no longer have to go to the hospital for checkups, and researchers can test volunteers on trials for new treatments faster and at a lower cost.
As of last Tuesday, the Parkinson’s Voice Initiative is 25% to their 10,000 target, so you should head over there and do your part, if you haven’t already. Place a simple three minute phone call and help the team build their voice sample database. For science!
(via New Scientist)
- Algorithm to tell why you’re smiling
- Mobile app to diagnose depression
- Medical breakthrough toward Ebola cure