From Code to Cure: AI Solutions for Modern Medicine

Artificial intelligence is transforming how we deliver healthcare and design new medical solutions. At the Faculty of IT, we’re harnessing that power through our new AI in Health initiative.

Led by Professor Enes Makalic, this interdisciplinary initiative brings together our best AI researchers with one goal: to solve real challenges across the entire health spectrum – from clinical medicine to pharmacy. Professor Makalic will introduce the initiative and share its new AI platform, built to accelerate discovery and create real-world impact. 

You’ll also hear directly from Monash researchers leading this work and discover four key projects that showcase the breadth of innovation already underway: 

• AI and cancer genomics

• Deep learning for biomolecular interactions

• Federated learning in digital health

• Multimodal and practice analytics for learning and performance in healthcare.

With the launch of Monash MAVERIC, this Monash Tech Talk is the moment to spark new collaborations across disciplines and industries. Join us, bring your ideas, and be part of shaping the future of healthcare.

Speakers and presentations

AI for smarter breast and brain cancer screening

Professor Enes Makalic

This presentation explores how AI is advancing cancer research and solutions in breast and brain cancers. For breast cancer, this talk will showcase research using machine learning to develop a novel, fully automated breast cancer predictor based on mammographic analysis. 

By combining mammographic density with textural image features, we have developed predictive models that generate personalised risk measures. Our findings demonstrate that these automated approaches predict breast cancer risk more accurately than conventional mammographic density measures. 

For brain cancer, this presentation introduces an Australian initiative to build the first national registry of families with multiple glioma cases, paving the way for better risk prediction and personalised care. Glioma is a rare and aggressive brain cancer with limited treatment options and a poorly understood genetic basis. 

Building on models from breast and prostate cancer research, the project will collect extensive clinical and biological data to uncover inherited risk factors, as well as integrate AI with genome-wide association studies and polygenic risk scores. Our recent work has already identified rare inherited variants that significantly contribute to glioma susceptibility, beyond known syndromes like Li-Fraumeni and Neurofibromatosis.

Deep learning for predicting biomolecular interactions

Professor Geoff Webb

Predicting how small molecules interact with proteins is a key challenge in drug discovery, especially when structural information is lacking. Existing approaches often require detailed protein structures and offer limited interpretability. 

We introduce PSICHIC, a graph neural network that efficiently learns the patterns of protein–molecule interactions using only sequence data. By integrating core physical and chemical constraints, PSICHIC achieves top-tier performance in predicting binding strength and provides interpretable insights into the mechanisms behind these interactions. 

Our findings demonstrate that sequence-based AI models can match—and even surpass—traditional structure-dependent methods, pointing to a new era of speed and insight for AI-driven drug discovery.

Federated learning in digital health

Dr Yasmeen George

Healthcare data is often siloed across institutions and jurisdictions, limiting opportunities to build robust AI/ML models for disease diagnosis and risk prediction. Centralised repositories promise scale but face significant legislative barriers to data sharing and privacy concerns, limiting access to this data. 

We build a federated learning platform that allows AI model learnings to be gained from health data across organisations and states without attempting traditional integration. By training AI models locally and iteratively across different sites, this platform preserves data privacy while scaling-up the knowledge.

From simulations to clinical practice: Multimodal and practice analytics for learning and performance in healthcare

Professor Dragan Gašević 

Healthcare education and practice generate vast amounts of data that can be harnessed to improve both professional learning and patient care. This talk presents recent work on two complementary strands of research. 

First, I will discuss multimodal learning analytics in nursing and healthcare simulations, where sensing technologies and artificial intelligence enable rich insights into learners’ actions, collaboration, and reflective practices. Our deployments in real-world nursing education settings reveal both the opportunities and the practical challenges of integrating these innovations into everyday teaching. 

Second, I will turn to practice analytics in healthcare, which make use of administrative and electronic health records to inform quality improvement and professional reflection. Studies in urology demonstrate how Bayesian modelling and trajectory clustering can support risk adjustment, identify patient subgroups, and detect outliers, while user interface innovations such as SeeCI show how clinicians can engage with repurposed data for meaningful practice reflection. 

Together, these two lines of research illustrate how educational and clinical data can be repurposed into actionable insights that support both learning and professional performance across the continuum of healthcare.