New James Cook University research has shown that AI can be used to help doctors differentiate between healthy people and people with schizophrenia, even when patients are stressed.
The research was published in the journal Biomedical Signal Processing and Control, led by JCU PhD candidate Mr Gideon Vos as part of a multidisciplinary team of engineers, data scientists, neuroscientists and psychology researchers.
Schizophrenia is a brain disorder that is reported to impact one percent of the world's population, with those suffering from the mental illness having high rates of mortality.
With most individuals experiencing symptoms before the onset of psychosis, improved diagnosis and early detection is crucial to its effective management.
Researchers tested new machine learning algorithms on the electroencephalography (EEG) brainwave patterns of healthy, stressed and schizophrenic patients, with AI results confirming the latest diagnostic observations from medical science.
EEG brainwave tests involve the detection of electrical impulses from different parts of a patient's brain during rest and while undertaking activities. But if experiencing stress, the brain responds differently for individuals with schizophrenia compared to healthy individuals.
"There are so many overlapping symptoms between acute stress response and people with schizophrenia," said Mr Vos.
"But there's something about the way that people with schizophrenia react to stress that is slightly different to what healthy people would experience under the same kind of situation."
Mr Vos explained that training an AI model with a large volume of high-quality data is a critical first step in the process of ensuring accurate, reliable and reproducible model predictions - essential requirements for AI's successful integration into the medical profession.
"A lot of machine learning models that are being built to predict schizophrenia are trained on data sets where they did EEG," he said.
"But EEG can be intrusive and uncomfortable, as it requires placing electrodes onto the patient's head, and you're in a lab rather than home environment, they're going to experience acute stress.
"It's important when building AI models that you understand the condition that they're building the model for."
The team of scientists used open-access EEG data sets and devised AI machine learning algorithms capable of accounting for the impact of stress on EEG brainwaves.
The stress-adjusted models yielded physiologically consistent patterns that aligned with established medical knowledge around schizophrenia diagnosis.
AI use in medical diagnosis does not aim for AI models to make the final decision - independent of medical professionals - but to instead use a process called 'explainable' AI to assist medical professionals to draw their own conclusions.
"If I train a model and the model says it can classify people perfectly into healthy and schizophrenic, most people would stop there, and say 'great, we've got a model that can predict schizophrenia'," said Mr Vos.
"But we need the AI model to explain why it can, or why it can't, separate these two groups.
"We need to be able to take those explanations and have a scientific rationale for why they occur … by the actual medical professionals."
Explainable AI provides a future pathway to more personalised and timely medical support for people in remote and regional communities, struggling to access medical services.
"I can go to my GP tomorrow to get a diagnosis … but a lot of people are not that lucky. They might be hours away from even just a basic GP service. Getting an expert opinion could take months." Mr Vos said.
"AI can be a tool on their smartphone that can see if something is potentially urgent, connecting them to a healthcare provider who can use the AI predictions and explanations in their own diagnostic process.
