Kylie Staats has always tried to live her best life. The intrepid traveller has skydived, rollerbladed, and bungee jumped while tearing across 42 countries, and she has even trekked the Great Wall of China. All while living with epilepsy. And while rollerblading is Kylie’s main form of transportation now, she dreams of one day of being able to drive a car and enjoy the freedom of the open road.
Fortunately for the 250,000 people in Australia who, like Kylie, are living with epilepsy, cloud technology is providing hope. Digital technologies are helping doctors and patients better understand the condition and make the diagnosis, treatment, and management of the neurological disorder faster and more effective, so that adventurers like Kylie can achieve her dreams and continue to live her best life.
The Implications of Epilepsy
Epilepsy costs Australia as much as AU$12.3 billion annually[1], and for those living with the condition, the journey from diagnosis to treatment can be painfully long and difficult. The complexity of analysing brain scans, genetic data, cognitive tests, and medical histories can make this journey a long one for patients. By leveraging cloud technology, the Australian Epilepsy Project (AEP) at The Florey Institute of Neuroscience and Mental Health[2], can help shorten this journey from diagnosis to management by up to 15 years - using cloud technology. For adventurers like Kylie, the gift of more independence is priceless.
The AEP was awarded $30m in MRFF funding in 2021, the largest single government investment into epilepsy research in Australia, to design and deliver a brain-health platform that will change the standard of care in epilepsy by solving the three core issues of access, uncertainty, and cost.
To accelerate the path to better epilepsy treatment, the AEP is turning to the power of artificial intelligence (AI) and cloud computing on Amazon Web Services (AWS). The project is cost-effectively building the world's largest connected dataset for epilepsy, to better understand the condition. For the connected data set. As of July 2024, the AEP has analysed data from more than 600 participants, and 300 control volunteers, with a goal of reaching 4,000 people by 2026. Participant and volunteer data is securely stored in Australia through cloud computing data centres, with researchers and collaborators only receiving de-identified information. And all data is encrypted during storage and transmission, with multifactor authentication protecting systems that access participant data.
Machine Learning to Prioritise Patients at Risk
This rapidly growing treasure trove of data on epilepsy includes: MRI scans that track brain activity during tasks like problem-solving and language; genetic tests with 600,000 data points analysed; medical histories of sufferers; and participant surveys. AWS cloud technology helps researchers rapidly correlate genetic similarities across participants and detect anomalies, while improving outcomes.
The vast knowledge gained from AEP’s multimodal dataset, which includes images, genetics, cognitive testing results, medical history, and other study data, gives clinicians a much-needed head start with diagnosing and treating epilepsy.
With AWS, insights from one ML model can then be funnelled to subsequent models, refining predictions for patient risk and recommended treatments. This means epilepsy researchers can integrate insights from multiple data sources, gaining a comprehensive view of each patient's condition and risk profile, ultimately facilitating more personalised care pathways.
Predictive Treatment with AI
The brain's complexity makes it challenging to detect patterns across different datasets. The standardised data gathered by AEP on AWS enables researchers to identify similarities across participants and compare them to the study’s control group.
With the ability to quickly and easily view patient details in a user-friendly and secure way, neurologists can help patients like Kylie better manage their condition to live a fuller life. AEP’s Medical History Explorer, enables the AEP team to search historical medical documentation for participants, targeting key study questions. It can reveal information like family history of epilepsy, the date of the first seizure, prescribed medications, and seizure symptoms.
Generative AI to Simplify Treatment
Generative artificial intelligence (generative AI) is helping too, rapidly surfacing relevant patient data and automating patient summaries to further accelerate and improve treatment. AEP also plans to build a generative AI chatbot for patients seeking general advice on epilepsy and treatment options, or the latest understanding from published scientific literature.
"The Australian Epilepsy Project demonstrates how healthcare organisations can leverage the secure, scalable computing resources of AWS to gain insights from complex data sets," said Simon Elisha, chief technologist for public sector in Australia and New Zealand for AWS. "With advanced cloud technologies, the AEP can efficiently process patient data and apply AI models to surface relevant clinical information and automate tasks. This enables neurologists to provide more timely and personalised care for those living with epilepsy."
“With AWS, we are building a world-leading epilepsy dataset, but the real impact comes from the insights we gain,” said Anton de Weger, AEP Digital and Technology Lead. “These insights enhance our understanding of epilepsy and brain function so we can help clinicians improve treatments. Integrating technology into our research project scales its impact, improves data analysis, and leads to better outcomes.”
The AEP leveraged the following AWS services to deliver better health outcomes:
- AEP uses Amazon Textract to extract text from various medical history document formats enabling searching and easy access to this valuable resource, which otherwise would be time consuming and difficult to analyse.
- A comprehensive MRI viewer has also been added to AEP’s portal hosted via Amazon EC2, allowing neurologists across Australia to quickly access high-resolution MRIs that are collected as part of the study without having to navigate multiple hospital systems and different viewers.
- Amazon QuickSight enables project data to be displayed via graphical dashboards.
- Amazon Elastic Kubernetes Service (Amazon EKS) helps AEP to efficiently run a variety of AI and ML workloads on data stored on AWS, providing the flexibility, scalability, and portability essential for their intricate epilepsy diagnosis and research workflows.
- The project uses Amazon SageMaker to develop and host various machine learning models that are used to extract and enhance the information captured in the multimodal project data.
- AEP used Amazon SageMaker Jumpstart to rapidly build a Patient Summary model that helps clinicians make enquiries in natural language.
Reference:
[1] Deloitte Access Economics. The economic burden of epilepsy in Australia, 2019- 2020 - Final report, February 2020
[2] The Australian Epilepsy Project receives funding from the Australian Government under the Medical Research Future Fund.
Also learn how AWS can help Australian organisations reduce carbon emissions of AI workloads.
