Computer scientists at the University of Rochester have developed an AI-powered, speech-based screening tool to detect signs of Parkinson’s disease.
It uses digital audio recordings to scan for subtle vocal patterns associated with the disease. The user simply records two short sentences into their computer's microphone, and artificial intelligence flags warning signs, if there are any.
"These acoustic models are trained on millions of recordings. They have access to a lot of the recordings available online, not specifically of people with Parkinson's, but of speech generally," said Abdelrahman Abdelkader, one of the two lead authors of a study on the tool.
After testing it, researchers said they found that the tool had an 85% accuracy rate.
Changes in the voice are one of the earliest signs of Parkinson's disease, according to the Parkinson's Foundation. They can include soft speech, monotone voice, slurring, and difficulty finding the right word. Limited facial expressions can also affect non-verbal communication.
The researchers' goal is to combine this tool with early screenings for other symptoms involving motor control.
The study's authors emphasize that their AI-based tool should not be used as a substitute for a clinical diagnosis. But they hope it will be an easy way for people, including those who live in remote areas, to detect potential changes that could warrant an evaluation by a neurologist.
The study was published in the journal NPJ Parkinson's Disease.
