In brief
Researchers have created an AI model, Delphi-2M, that analyzes medical records to predict the likelihood of over 1,200 diseases. This innovative technology acts similarly to weather forecasts, identifying high-risk individuals and helping healthcare systems prepare for future demand. Initial testing shows promise, but further refinement and ethical considerations are necessary before clinical use.
Scientists have unveiled a groundbreaking artificial intelligence (AI) model, Delphi-2M, which can predict health problems as much as ten years into the future. By identifying patterns in extensive medical records, the model forecasts the risk of over 1,200 diseases, akin to a weather forecast that estimates rain probabilities. Researchers envision utilizing this technology for proactive healthcare measures, such as identifying high-risk patients, which could lead to interventions aimed at preventing diseases before they manifest.The Delphi-2M model draws upon the same principles that underlie AI chatbots like ChatGPT, focusing on pattern recognition within anonymized medical data. It does not provide specific dates for potential health events but rather calculates the likelihood of various diseases occurring. Professor Ewan Birney, interim executive director of the European Molecular Biology Laboratory, highlighted the model's ability to conduct simultaneous probability assessments across numerous diseases—a feat previously unachieved in the medical field.The model was initially trained using anonymized UK Biobank data, which encompasses hospital admissions, GP records, and lifestyle choices from more than 400,000 participants. After validation through additional datasets, including medical records from Denmark, the model's predictions have proven reliable, particularly for diseases like type 2 diabetes and heart attacks that show clear patterns over time.Currently, while the AI tool is not ready for clinical implementation, researchers intend to employ it for early patient identification and tailoring preventive strategies—such as lifestyle modifications based on individual risk profiles. Ultimately, there are aspirations for the model to assist healthcare providers in anticipating resource demands in specific regions, enhancing the overall planning of healthcare systems.However, experts caution that further research and testing are needed, especially to mitigate biases present due to data limitations. The model's development represents a significant step towards integrating scalable and ethically responsible predictive modeling into medicine, according to Professor Gustavo Sudre of King’s College London. Collaboration among institutions like the European Molecular Biology Laboratory and the German Cancer Research Centre aims to ensure this technology paves the way for a new frontier in healthcare.
