How machine learning is helping solve health and medical challenges

26 Feb 2024

Recently appointed QDHeC’s HeRA researcher for Federated Learning (FL), he will be working with colleagues at UQ and globally to maximise the research and clinical impact of health data, both through FL - a machine-learning paradigm where partners collaborate on complex research without centralising or sharing data outside of their organisations – and the generation of readily-accessible, highly realistic synthetic health data.

“Federated Learning will enable us to develop models by bringing the models to the data at various healthcare institutions rather than sending the data out,” Sebastiano said.

He is excited about QDHeC’s focus on synthetic data - computer-generated medical records for testing and training AI models with similar accuracy to real data.

“The use of synthetic data reduces the privacy risks which surround sharing real patient data for health research. Using synthetic data will accelerate how we can work on solving some of the big health challenges of our time,” Sebastiano said.

Sebastiano, who completed his PhD in computer science at the Fraunhofer MeVis Institute for Digital Medicine and Jacobs University Bremen, Germany, has spent the last six years at UNSW’s Centre for Big Data Research in Health working on novel machine learning methods and applying these techniques to various health and medical challenges.

“One major field that I am working on with colleagues from Australia and New Zealand is on the use of machine learning methods to predict cardiovascular risk in the general population with higher accuracy than traditional statistical methods,” he said.

Looking ahead, Sebastiano believes that developing robust governance models for the use of synthetic data and other AI applications in health and medicine is essential.

“The models around the use of synthetic data are constantly improving, they are becoming more accurate, essentially synthetic data is by now very similar to the real data.

“I think what we need to do is to establish the appropriate governance structures around the use of all of this data and consumers will play an important role in helping develop the guidelines.”

For his PhD, Sebastiano worked on diffusion tensor MRI imaging for neurosurgery planning: “I was fascinated by the mathematical aspect of this research because we were essentially working with six-dimensional images. There was also this very useful, practical aspect, in helping surgeons plan their surgeries with as much accuracy as possible.”

“I guess I really liked doing work that was both mathematically challenging and had a positive impact on patients. So I decided to stay in this field to discover other ways in which AI and machine learning can be used to help patients."