Skip to main content

Thuy applies artificial intelligence to speech pathology

Adjunct Associate Professor Thuy Frakking.

Artificial intelligence (AI) could soon diagnose swallowing difficulties in critically unwell children using sound, reducing the need for costly, radiation-exposing x-rays and making diagnosis more accessible for clinicians worldwide.

Adjunct Associate Professor Thuy Frakking, a recipient of an inaugural Gold Coast Health Clinician Researcher Fellowship, will spend the next three years collecting swallowing sounds in our children’s intensive care unit to train an AI algorithm to recognise swallowing difficulties.

Working alongside computer engineers at Griffith University, A/Prof Frakking has already demonstrated that machine learning can recognise a swallowing sound in a child and classify the sound as impaired or not impaired.

The next phase of her research will focus on proving the algorithm can work in the children’s intensive care space, where around half of children will experience a swallowing difficulty.

“If we can diagnose swallowing impairment at an early stage, we hope that we can recommence children safely eating and drinking again sooner. Doing so will improve the trajectory of their hospital stay,” says A/Prof Frakking, an advanced speech pathologist.

Speech pathologists already diagnose swallowing impairment by listening through a stethoscope. A/Prof Frakking said one third of speech pathologists do so, but it is a difficult technique to learn and takes experience to master.

“We can make it more objective by using machine learning, because my interpretation of swallow and breath sounds can be quite subjective,” A/Prof Frakking says.

“If my machine learning is accurate and objective enough, clinicians can use it and be confident that there is or is not a swallowing impairment, and they don’t need to refer on for an x-ray swallow.”

The x-rays that diagnose swallowing difficulties are called videofluoroscopic swallowing studies; they’re expensive, labour intensive, expose patients to radiation and have a high failure rate in children.

Patients must consume barium to take part in the study, which is often the obstacle when working with children – many refuse to eat the yogurt-barium mix.

“Will we need a videoflouroscopic swallow studies in the future? I hope not,” A/Prof Frakking says of the potential of AI.

The long-term goal for A/Prof Frakking is to commercialise an app that is easily accessible worldwide, improving care for unwell children everywhere, as many areas of the world do not have access to videoflourosopic swallow studies.

This technology could also be useful for new mothers, as the algorithm could be trained to recognise when a baby is feeding and also measure the volume of milk the baby swallows.

A/Prof Frakking says she is ‘forever grateful’ for receiving an inaugural Clinician Researcher Fellowship, which allows her two days per week dedicated research time for three years thanks to her Clinician Researcher Fellowship.

In addition to working on artificial intelligence, A/Prof Frakking will also mentor up-and-coming Gold Coast Health clinician-researchers.

“This scheme will build a bank of leading world-class clinician researchers embedded in Gold Coast Health in the long term and it will improve Gold Coast Health’s research reputation locally and internationally. It will help nuture PhD and junior clinicians into the clinician researcher pathway,” she says.

Applications are now open for the 2024 round of Clinician Researcher Fellowships. Find more information here.
 


Last updated 25 Oct 2024