Statistical Society of Australia J.B. Douglas Postgraduate Award and International Biometric Conference Representative

Using machine learning to help stroke victims

When Tea Uggen left her home in Norway to take up an undergraduate degree at the University of Technology Sydney, she wasn't quite sure where she was headed. She was interested in linguistics and psychology, but ultimately followed her family's practical advice and enrolled in a business degree.

It was a compulsory statistics subject in her first semester that put Tea on the road to the PhD in applied statistics she has now almost completed, but not in the way you might expect.

"I was struggling a bit with the language and the culture and with being away from everyone I knew," Tea says. "And I found statistics especially difficult. I grew to hate it because I didn't understand anything. I failed."

Tea went to see a course adviser for help, and when she repeated the subject she was better prepared. "I went to every class and every help class, and really studied properly."

Second time around, Tea not only passed with a distinction but discovered a deep interest in the subject. Tea pursued econometrics – economic statistics – through the rest of her business degree before embarking on a Master of Mathematical and Statistical Modelling.

This was where she found a project that brought together her interests in language, health and statistics. Associate Professor Tapan Rai, who is now Tea's supervisor, was using statistical methods to study a specific language impairment called aphasia in people who have had strokes.

Tea's project grew into a PhD, and she is now using machine learning to develop an automated way to determine the severity of a patient's aphasia to make sure they get the most appropriate treatment.

Currently, diagnosing the level of severity is a long manual process – a speech pathologist works through a lengthy assessment with the patient, and scoring the results can take even longer. Tea's work shows severity can be accurately determined by analysing transcripts from a five-minute test.

Tea aims to submit her thesis in the first half of 2019. And after that? "I'd like to keep working with people with aphasia, other communication disorders or mental health issues. It's important to me that my research is applied to help people."