UNSW Sydney scientists are working to develop Australia's first commercialised lipid-based blood test for breast cancer detection powered by artificial intelligence (AI) – with a version of it already being used in clinics.
Associate Professor Fatemeh Vafaee and her team at the Vafaee Lab envision developing AI-enabled blood testing platforms to detect and monitor a range of cancers such as lung, liver and brain tumours.
"The way cancer detection works now is that it often only picks up the tumour when it is already established," A/Prof. Vafaee says.
"The method of detection itself relies on imaging and invasive tissue biopsies, which carry their own risks and may miss parts of the tumour," she says.
A/Prof. Vafaee says tumours have many types of cancer cells, all with different characteristics and behaviours.
"So, depending on where the needle goes in, we're always going to miss the entire population that exists for that tumour."
She says enhancing the ability of simple blood tests to detect tumour initiation before symptoms appear allows for earlier intervention and improved outcomes.
As tumours rely on blood for their nutrition and growth, "blood is continuously communicating with tumours, releasing biomarkers such as DNA, RNA, proteins and metabolites throughout the body," says A/Prof. Vafaee.
"These biomarkers can be used to detect cancer long before a lump is visible in imaging – in principle, months or years before. And with minimal pain and risk."
How does AI help?
Detecting cancer is not the only application of AI-driven blood tests – they can also help monitor cancer progression, as well as treatment responses, in real-time.
Far beyond its most common application – powering large language models such as ChatGPT – AI can be used to advance many other applications, including in medicine. In the Vafaee Lab it is used to analyse large datasets across an array of biomarkers.
These biomarkers come with their own set of data complexity – as well as a lack of reproducibility.
"There are hundreds of thousands to millions of molecules in one individual sample," A/Prof. Vafaee says. "So identifying the subtle molecular patterns indicative of cancer from all these data points is like searching for a needle in a haystack."
In her lab, she applies AI to improve and speed up the process through advanced algorithms that can sift through complex data, identifying meaningful molecular signals in blood that are predictive of tumour initiation or progression. This is an improvement on current detection methods, allowing for personalised approaches as well as reproducible results.
"Importantly, it also allows us to uncover the 'why' behind the AI's predictions," says Associate Professor Vafaee. "By integrating explainable AI techniques, we ensure that the models provide not only accurate outcomes but also clinically interpretable insights—crucial for building trust and supporting decision-making in real-world healthcare settings."
One lab, a world of opportunities
A/Prof. Vafaee's lab is a globally unique setting for AI-powered cancer studies.
She and her team combine a range of data – from medical images to text-based electronic health records to molecular data – to build a 'big picture' of health and disease.
Along with Australian biotechnology company BCAL Diagnostics, they have already developed AI algorithms for a blood test that offers a non-invasive way to rule out breast cancer.
The test successfully transitioned from the research phase to clinical application in March 2025 and is now available in multiple specialist breast clinics across Sydney and Melbourne. The test enhances early detection particularly in women with dense breast tissue, where traditional imaging methods may be less effective.
The Vafaee Lab is part of both a national and international consortium working to integrate blood tests into existing national breast screening programs, which will help improve detection across more populations.
A/Prof. Vafaee says while challenges remain in integrating this type of technology at scale, AI is poised to reshape the realm of personalised medicine.
She is leading a large-scale study for 'multi-analyte' blood tests, which analyse a combination of biological markers —proteins, metabolite, RNAs — instead of relying on a single indicator. This approach significantly improves the sensitivity and specificity of cancer detection.
"The challenge remains that we are trying to integrate diverse data types into a cohesive system for a holistic view of a patient's health," A/Prof. Vafaee says.
"We are highly adaptive in applying a range of methods from traditional machine learning to cutting-edge AI techniques.
"We're also looking to create tests based on other biofluids such as urine or saliva."
She says as the complexity of the AI models grows, she and her team are also working to ensure these are interpretable and accepted by the medical community.
"We can use AI not only to detect cancer early but to improve the lives of those who have it."
Key Facts:
A UNSW scientist is working to bring AI-powered blood tests for cancer detection to market – offering earlier diagnoses, reducing the need for invasive procedures and paving the way for more personalised treatments for all.
