A study led by Australia's University of Melbourne has developed an artificial intelligence (AI) tool capable of creating a "digital twin," or virtual replica of a patient, to predict an individual's health trajectory.
The research team used three datasets containing thousands of patients' electronic health records to train a large existing language model, according to a statement from the University of Melbourne .
The AI model, dubbed DT-GPT, which has been hailed as a "potential gamechanger for the clinical trials sector," analyzed medical data from patients with Alzheimer's disease or non-small cell lung cancer, as well as those admitted to intensive care units (ICUs).
The model created a digital twin of the patient and accurately predicted changes in their health over time by leveraging existing medical knowledge and analyzing patient histories, including laboratory results, diagnoses, and treatments, the statement added.
The model was blinded to patient outcomes, allowing the research team to validate its predictions. The statement also noted that the DT-GPT model outperformed 14 other state-of-the-art machine learning models in predictive accuracy.
"This technology paves the way for a shift from reactive to predictive and personalized medicine," said University of Melbourne Associate Professor Michael Menden, lead author of the study published in the journal NPJ Digital Medicine.
"The model allows doctors to anticipate when their patient's health will deteriorate so they can intervene earlier," Menden said, adding that the model can also predict drug side effects, helping doctors tailor treatment to a patient's unique profile. each patient, and improve health outcomes.
The model quickly interprets complex data and has a chatbot-like interface for users to explore its predictions, Menden said, adding that DT-GPT uses generative AI to create "zero-shot predictions," meaning the model can make estimates based on data about laboratory values even without prior training.