Digital twin technology drives new era in predictive medicine
Predictive technology from Melbourne scientists offers a new way for clinicians to anticipate deterioration and tailor treatment using digital patient replicas.
A new AI model capable of generating digital twins of patients is being hailed as a significant step forward for clinical research. Developed at the University of Melbourne, the system reviews health records to predict how a patient’s condition may change during treatment.
DT-GPT, the model in question, was trained on thousands of records covering Alzheimer’s disease, non-small cell lung cancer, and intensive care admissions. Researchers stated that the model accurately predicted shifts in key clinical indicators, utilising medical literature and patient histories.
Predictions were validated without giving DT-GPT access to actual outcomes, strengthening confidence in its performance.
Lead researcher Associate Professor Michael Menden said the tool not only replicated patient profiles but also outperformed fourteen advanced machine-learning systems.
The ability to simulate clinical trial outcomes could lower costs and accelerate drug development, while enabling clinicians to anticipate deterioration and tailor treatment plans more effectively.
Researchers also noted that DT-GPT’s zero-shot ability to predict medical values it had never been trained on. The team has formed a company with the Royal Melbourne Women’s Hospital to apply the technology to patients with endometriosis, demonstrating wider potential in healthcare.
Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!
