Older adults will one day be able to talk into a smartphone to test for dementia, should work on a new speech app prove fruitful.
Dr Beena Ahmed from UNSW’s School of Electrical Engineering and Telecommunications is leading work on an app that uses machine learning technology to look at paralinguistic features of a person’s speech, such as prosody, pitch, volume and intonation, as well as testing memory recall.
Ahmed said the tool would replace current subjective procedures that “have limited diagnostic accuracy”, often depending on the expertise of the clinician and test conditions, among other factors.
“Also, access to specialist services is challenging and often inequitable, especially in rural and remote areas,” Ahmed said. “Speech, on the other hand is easy to collect, even remotely over the phone, is cost-effective and non-invasive. And as our speech-based assessment will be performed computationally, it is less susceptible to subjective biases.”
Users of the app would listen to a list of 15 words and then be asked to repeat back as many as they could remember. The process would be repeated three times using the same 15 words, then after a 20-minute break, the person would be asked to recall as many words as they can.
But the app wouldn’t only be looking out for accurate word recall, but other signs of dementia, including frequent pauses in searching for a word, repeated or restarted phrases, repeated or extended syllables, frequent fillers, repaired utterances, mispronunciations, word substitutions, as well as certain effects in the speaker’s melody, intonation and rhythm.
And Ahmed said early results appeared promising.
“Initial studies by our team have shown that is it possible to discriminate between participants at high risk and low risk of dementia as determined by clinicians with an accuracy of 94.7 per cent when trained with paralinguistic features only, and 97.2 per cent when trained with paralinguistic and episodic memory features using audio recordings of participants completing an episodic memory test.”
The app is still in its early stages and could be available for use in five to 10 years.Do you have an idea for a story?
Email [email protected]