Machines learn to pick up the cues

22 October 2014

A new data pattern recognition program is pinpointing people most at risk of suicide.

Deakin researchers have developed an artificial intelligence program that can more accurately predict the mental health patients most at risk of suicide.

Professor Svetha Venkatesh, Director of the Centre for Pattern Recognition and Data Analytics (PRaDA), and her team have been working with Barwon Health on the project for over two years. Dr Truyen Tran was the research lead.

Using hospital electronic records for patients with a mental illness, their program is able to analyse large of chunks of data and is three times more accurate than previous risk detection systems.

It has been estimated that between 80-90 per cent of people who commit suicide have a mental illness.

“Our Health Analytics program was able to take into account extensive data from Electronic Medical Records for each of the 9000 patients from the mental health cohort of Barwon Health,” said Professor Venkatesh. “From this data, it identified patterns that showed those most at risk.”

“In contrast, the manual risk assessments that are usually made by medical practitioners consider around 17 factors, categorising patients as either high, medium or low risk. Until now, it would have been impossible to analyse this much data and discover patterns in the risk factors.”

The factors considered ranged from the patient history of emergency attendance, to the types of injuries they were admitted for.

The PRaDA program is now being tested by Barwon Health, providing a complete risk profile for each patient with a mental illness and enabling practitioners to view the risk-relevant patient information more effectively.

“It is allowing better resource utilisation, providing support where it’s needed most and it should, hopefully, help to reduce the incidence of suicide in the region,” said Professor Venkatesh.

She added that the program could be adopted by health systems across the country and overseas.

PRaDA is one of Australia’s leading research centres in the field of machine learning and pattern recognition, with researchers developing programs that analyse large-scale data patterns in areas as diverse as surveillance, social media and health.

In fact, machine learning is now the most popular course at Stanford - and is being hailed as the key to technological transformation across many industries. It is a type of artificial intelligence that enables computers to learn without being explicitly programmed, so they can grow when exposed to new data, and is now used in Google, Bing, Facebook, computer vision and spam.

At PRaDA, researchers have used machine learning to develop the award-winning Toby Playpad app, which is an adaptive early intervention program for children with autism, monitoring each child’s performance and adjusting lessons accordingly.

The Centre is also currently working with the Black Dog Institute to identify deviations in social media that could predict suicide risk and help with the development of an early warning system.

“Now that we can analyse such vast amounts of data, there is enormous potential to gauge the mental health of whole populations and develop much more effective public health campaigns,” said Professor Venkatesh.

People wishing to discuss suicide or depression can contact "Beyond Blue" on tel: 1300 22 4636, or contact the Suicide Call Back Service at www.suicidecallbackservice.org.au

Professor Svetha Venkatesh (top) and Dr Truyen Tran. Professor Svetha Venkatesh (top) and Dr Truyen Tran.

More like this

Research news