A new app that records and analyses people's voices could help with the early detection of Parkinson's disease, one of the team behind its development says.
Professor Dinesh Kumar, an engineer from the Royal Melbourne Institute of Technology, started the study which led to the app's development seven years ago.
He told Afternoons with Jesse Mulligan that by analysing how a person said three different phonemes - A, O and M - the app could detect whether someone might have the degenerative disorder in just 10 seconds.
Photo: 123RF
Currently, a Parkinson's disease diagnosis requires a 90-minute evaluation with a neurologist.
Kumar said the three sounds people were asked to speak were chosen because their production corresponded to three different parts of the "vocal model", namely the throat, the mouth and the nasal cavity.
The intention of the app was "to recognise the very small changes that happen in the voices of people because of the disease", he said.
"We're not looking at actually the differences between two voices, but how do people create the three different sounds."
Once recorded, the team's "AI machine" could identify whether there was a link between a voice and Parkinson's disease, or not, he said.
The app's accuracy in identifying Parkinson's disease was between 82 and 90 percent, Kumar said, and the technology it used had also been found to have other uses.
One of Kumar's colleagues found himself stuck in Indonesia during the Covid-19 pandemic, where he had access to patients with Covid-19.
Using a different model but the same approach, the team used the technology to see whether they could determine a difference between those with Covid-19 who required hospitalisation and those who didn't.
"We found that we could detect the difference between the two very well."
Kumar said while Parkinson's was not currently curable, it could be managed if detected early, which was where he hoped the app would be of use.
"Many times people delay getting their diagnosis done because many of the symptoms do not appear until late, so there is an urgent need to detect the disease early."
The app was not, however designed to be used by members of the public, he said.
"We do not want you to use the app - this app is to assist a neurologist or a clinician to be able to first monitor their patients and to perform screening."
Kumar said he hoped that, in time, it could be used as part of regular GP check-ups - in the same way screening for cholesterol was currently undertaken - and that it would help clinicians decide whether a patient needed to be referred to a neurologist or not.
He could also see the potential for the technology to be used in places around the world that had fewer diagnostic resources.
"Smartphones are available everywhere and this is something that does not need any high tech, it can be done quickly, we are hoping it is something that can be rolled [out] around the globe."
The next step would be to have the app certified by the Therapeutic Goods Administration (TGA) - the authority charged with regulating evaluating, assessing and monitoring products that are defined as therapeutic goods in Australia, Kumar said.
"We are technically ready, we've got the validation done, but in order to convince the TGA that we've got everything in place, we would need to conduct more tests, as well as then look for partners in the medical world who would be willing to try it out."
