AI Impact Projects

This project uses highly advanced AI techniques to accelerate drug This includes using meta-learning (MAML) for “few-shot learning” (creating accurate models from very little data) and using contrastive learning for powerful “zero-shot” prediction of promising drug compounds for entirely new protein targets.

Phase-contrast CT is a new imaging technique that provides higher resolution images of soft tissues but requires a high radiation This project will develop an AI model to “denoise” images captured at lower doses, making them clinically useful.

This project uses AI to enhance vaccine safety monitoring by analysing millions of public social media conversations in real-time. Researchers applied Natural Language Processing (NLP) to three million public tweets to automatically identify personal health mentions related to vaccine side effects during the COVID-19 pandemic.

This project will use deep learning models to find correlations between distinct imaging phenotypes in routine PET/CT scans and the underlying multi-omics signatures of a patient’s tumour, as well as their response to treatment.

A national AI-enabled brain-health platform translating 30+ years of Florey research in imaging, genetics and psychology into patient outcomes, already connecting >1,500 patients with >160 neurologists across all mainland states. Platform components include a clinician web portal with GenAI patient summaries and ingestion of unstructured medical history, an automated ~20-hour imaging pipeline, ML analysis (e.g., activation analysis), a custom tele-neuropsychology tool (TENT), and a genetics pipeline—hosted on AWS with HPC integration.

Baby Moves is a clinical research project that uses AI to analyse smartphone videos of infant movements, supporting the early and accessible detection of cerebral palsy. The system applies deep learning-based pose estimation to home-recorded videos to identify abnormal movements that are early indicators of motor development issues.

Applies a range of ML and deep learning approaches (including Transformers) to multi-omics data to define molecular subtypes and identify biomarkers in several cancer types, including prostate cancer, melanoma and brain metastasis.These groundbreaking projects serve as beacons, illuminating both our tremendous potential and the challenges we face. They offer a dual perspective:

A major international collaboration between MCRI and the Gladstone Institutes (San Francisco) that uses patient-derived stem cells to grow 3D models of human heart tissue. By combining this platform with high-throughput experiments, genomics, and AI, researchers seek to uncover the fundamental causes of congenital heart disease and develop novel precision treatments.

Image: Proliferation (blue) of heart muscle cells (red) in a stem cell derived 2D cardiac model.

This project uses deep learning and computer vision to extrapolate 2D spatial-omics information from tissue sections into faithful 3D in silico representations of tumours.

A programme of simulation and in-situ evaluation to measure workflow, safety and efficiency impacts of “ambient” AI scribe systems that capture clinician–patient encounters, transcribe audio, and use large language models to generate clinical outputs (notes, referrals, forms) across metropolitan and rural Victorian sites.

This project aims to develop a novel AI method for the rapid and accurate taxonomic classification of bacteria from genomic It will use Geometric Deep Learning to represent complex relationships within bacterial genomes as a graph, allowing the model to learn structural features missed by traditional methods.

This project combines AI-based analysis of medical imaging (CT, DEXA, OCT) with genomics and other omics data to identify the molecular drivers of phenotypic variation observed in the images.

A new Florey–ACAMI collaboration to develop an AI model for CNS-targeted mRNA therapeutics, integrating advanced AI with neurobiology to improve mRNA stability/translation and LNP targeting/transfection for dementia and related neurodegenerative diseases.

Combines public AI tools like RFdiffusion and AlphaFold into a portable NextFlow pipeline to design novel proteins, which will then be validated and produced in WEHI’s Protein Production Facility.

The retina serves as a window to systemic health (oculomics). This project aims to leverage AI-driven oculomics to study multi-system diseases, particularly targeting early detection.

A large-scale, MRFF-funded initiative to establish an integrated, high-capacity human phenotyping The project will process, store, and analyse multi-omics data and deep phenotype information (e.g., cognition, mental health) to deliver personalised medicine at scale.

AASGARD is a newly funded, $8 million national initiative designed to harness AI to improve the diagnosis of severe and rare genetic diseases in children. Supported by the MRFF, the project will develop, test, and validate AI-powered tools to more accurately and safely identify disease-causing variants in genomic data.