With a quick phone call to a computer, a new voice algorithm could diagnose diseases by using the sound of your voice.
The software uses a speech-processing algorithm to identify telltale changes in the voice of a person with Parkinson’s disease, New Scientist explains.
Although surgery and drugs can hold back the disease’s progression, there’s no cure. Diagnosing it relies on an expensive and time-consuming assessment.
In most sufferers, the disease manifests early on in the voice, affecting the ability to control the vocal cords and soft palate. Common signs include: a tremble in the voice, softer speech, and breathiness or hoarseness — which can all be subtle in the beginning.
The voice-based system — developed by Max Little of Massachusetts Institute of Technology’s Media Lab — was used to process 263 recordings of 43 people who were asked to sustain 6 or 7 vowel “ah” sounds.
The algorithm was trained on 10 impairments (or dysphonias) in these recordings. It managed to diagnose Parkinson’s speech markers with an 86% accuracy; that jumps to 99% with people in later stages.
The Parkinson’s Voice Initiative website lists phone numbers for people to call the computerized diagnostic system. Little wants people around the world to test his system. By getting non-sufferers to call in, the system can learn to weed out and discard unnecessary voice patterns, such as those brought on by a cold or heavy smoking, Wired explains.
In the future, data could be collected using Intel’s At-Home Testing Device, a telemonitoring system. It would then be sent to a clinic where the algorithm processes it and maps out the speech, identifying markers on the Unified Parkinson’s Disease Rating Scale (UPDRS) so that the severity of the illness is known.
The technology could be available to doctors within the next couple years.
The work was presented at the TEDglobal conference in Edinburgh last month. A paper [pdf] was published in the IEEE Transactions earlier this year.
[Via New Scientist, Wired, BBC]
Image: Parkinson’s Voice Initiative
