An AI powered selection game where users click either the yellow option or green option to collect rewards (smiley faces).
These are collated at the bottom of the screen.
Users have 100 'clicks' to collect the most rewards.
Announced today at D61+ LIVE in Sydney, the researchers revealed that using a simple computer game and artificial intelligence techniques, they were able to identify behavioural patterns in subjects with depression and bipolar disorder, down to subtle individual differences in each group.
The study included 101 participants: 34 with depression, 33 with bipolar disorder, and a control group of 34 subjects.
The computer game presents individuals with two choices, and tracks their behaviour as they respond.
The complex data collected from the game is analysed through artificial neural networks — brain-inspired systems intended to replicate the way that humans learn — which are able to disentangle the nuanced behavioural differences between healthy individuals, and those with depression or bipolar disorder.
Dr Amir Dezfouli, lead author of the research, a neuroscientist and machine learning expert at CSIRO’s Data61 said the research represented a possible step-change in the emerging field of computational psychiatry.
“Currently 69 per cent of bipolar patients are initially misdiagnosed, and around one-third of these patients might remain misdiagnosed for 10 years or more,” Dr Dezfouli said.
“If we can understand how the brain works, we can develop more accurate processes for diagnosis and more effective treatments for people with mental health disorders.
“Artificial intelligence and deep learning techniques allow us to analyse complex datasets and make accurate models of the brain processes involved in psychiatric disorders.
“Characterising mental health disorders in granular detail could allow clinicians to develop more personalised treatment plans based on an individual’s unique diagnosis,” Dr Dezfouli said.
Clinicians around the world mainly rely on the Diagnostic and Statistical Manual (DSM) and the International Classification of Diseases (ICD), which provide a set of criteria for the diagnosis of mental health disorders.
“We hope that this technique will build on the DSM and provide an additional decision-making tool for clinicians,” Dr Dezfouli said.
“The strength of the computer game is that unlike traditional mental health assessments, the results can directly reflect the brain processes that are affected due to the disorders, as individuals are responding to stimuli rather than direct questions about their mental state.”
Dr Richard Nock, machine learning group leader at CSIRO's Data61 said artificial intelligence is a powerful general purpose technology with the potential to help solve some of the greatest challenges relating to health, energy, ageing, safety, security, climate and the environment.
“While artificial intelligence holds enormous potential it must be deployed with privacy, ethics and inclusiveness at its core,” Dr Nock said.
“We need to design systems that deliver benefits individually and collectively. The artificial neural network was specifically designed to produce interpretable results, and will augment the capabilities of clinicians and psychiatrists.”
The researchers are looking to partner with hospitals and mental health research centres to conduct further research to validate the technique for real-world use, providing decision support for clinicians.
The research paper, Disentangled behavioral representations, has been accepted at the 2019 Conference on Neural Information Processing Systems (NeurIPS), the world’s largest machine learning conference taking place in Vancouver in December.
Images
B-roll video
Background information
About the research
The research paper, Disentangled behavioral representations, was co-authored by Amir Dezfouli, Hassan Ashtiani, Omar Ghattas, Richard Nock, Peter Dayan and Cheng Soon Ong. The research is a collaborative work by CSIRO’s Data61, McMaster University, University of Chicago, Australian National University, University of Sydney and Max Planck Institute.
The work was made possible through the data provided by Professor Bernard. W. Balleine at the University of New South Wales and Dr Kristi Griffiths at the University of Sydney.