Case Study: How the Australian Epilepsy Project taps tech to empower patients – Digital Nation – Emerging Tech

The Australian Epilepsy Project (AEP) aims to build a dataset on epilepsy and tap AI and cloud to help industry players better understand the condition.

Both chief investigator Graeme Jackson and digital and technology lead Anton de Weger told Digital Nation that as the world’s largest connected dataset, it would improve healthcare workers’ understanding of epilepsy and help patients better manage it.

“The motivation is to change the lives of patients. One in 50 people will have active epilepsy. Up to two percent of the population [and] more than 10 percent will have a seizure at some time in their life,” Jackson said.

“As a clinician and researcher, I want solutions at population scale.

“We’re on the cusp of [having] the ability to take the expertise that lives in a bricks-and-mortar hospital out through data to population scale.

“You can’t make a thousand epileptologists, but you can scale data trained by us and survey that population [so] everyone gets optimal care early in their disease.

“The healthcare consequences and the health economic consequences are enormous.”

The project forms part of The Florey Institute of Neuroscience and Mental Health and is also connected to the University of Melbourne and Austin Health.

The team’s work has led to a portal for clinicians to centralise reporting and machine learning is also being used to understand unstructured data.

“We went to AWS and decided to use them to provide all the hosting services and a lot of the other services that they provide around health care,” de Weger said.

“We use that as a basis for our system.

“We’ve developed a clinician’s portal which allows a neurologist to log in and view a standardised report, so we can deal with security and making sure that all of the information stays where it’s supposed to be.

“All of the data stays in the Australian environment. It’s all encrypted at rest,” de Weger said.

He added from here the team has started to further use AWS services to assist with analysing its data and machine learning models, and use capabilities around understanding unstructured data.

“For each of our patients, we’ve got information in their medical history, scans of faxes, letters, Word documents, PDFs.”

The team is using AWS Textract to convert this unstructured information into text and “large language models to understand what it’s talking about in each of those sections”.

“We use a retrieval-augmented generation process where we ask the large language model to answer questions on that text based off the information that we provided,” de Weger said. 

Jackson said that as a doctor, the platform has been a turning point.

“We’re building a modern healthcare platform connecting data in a standardised way, as a new standard of care for researchers so they can understand the disease and find treatments because nothing is static”. 

He said the tools have been designed with clinicians in mind. Healthcare workers played a key role in the tech development, providing “guidance for how to design our products and where to focus their attention”.

“When I started two years ago, we had a couple of hundred participants and we’re trying to grow to 4000 participants. That size of a connected data set in epilepsy, would be one of the largest in the world with high resolution capability.

Patients are benefiting the most from access to insights that enable them to stay informed of their care.

Building scale

Next steps for the project include scaling and expanding into the paediatric space and continuing to build out insights.

De Weger said the project is starting to find correlations and patterns in the data, “which we think will build better insights and provide information to the neurologist.”

“We’re building a way of processing data around a complex organ like the brain that we could standardise and use to learn more about the brain and how the brain functions in other conditions.”

The work is also seeking to expand to cover children as well.

Meanwhile, de Weger said the plan is not to replace clinicians but use technology to gain “good quality outcomes for everybody that’s standardised”.

“That is how the future of how AI and technology can help in healthcare – fixing the protocol at the beginning for a period of time, so that you can build sequential small machine learning models and algorithms to enhance data and give a standardised output.”