An AI-powered Learning Health System platform for Hospital Surveillance of invasive fungal infections

Opportunity

Invasive fungal infections are rare and difficult to detect, yet they can have a serious detrimental effect on immunocompromised cancer patients. Timely, accurate surveillance is critical – but existing approaches rely heavily on manual review, limiting scalability and responsiveness.

Intervention

The Centre’s research team developed secure data infrastructure drawing on hospital electronic health records, and applied AI to process free-text pathology and imaging reports to automatically detect infection. An industry-grade continuous integration and continuous delivery (CI/CD) data pipeline was also developed to support ongoing algorithm refinement, with end-user feedback collected throughout to drive validation and optimisation.

Impact of simulation

The platform achieved automated infection detection using multimodal electronic health record data with 91% sensitivity, alongside new tools to analyse free-text reports for signs of infection.

The project also established robust infrastructure for ongoing algorithm validation and refinement, providing a foundation for accurate, safe, and trustworthy clinical AI in hospital surveillance settings.