Delphi-2M is a generative AI model developed by researchers from the European Molecular Biology Laboratory (EMBL), the German Cancer Research Centre, and the University of Copenhagen that can predict the risk of over 1,000 diseases up to 20 years in advance by analyzing an individual's medical history and lifestyle factors. The model, described in a study published in Nature on September 17, 2025, uses a modified GPT architecture trained on anonymized health data from 400,000 UK Biobank participants and validated on 1.9 million Danish patient records, demonstrating high accuracy comparable to established single-disease risk models.
Delphi-2M analyzes sequences of medical diagnoses, demographics (age, sex), and lifestyle data (smoking, alcohol, BMI) to forecast future health trajectories, estimating the likelihood of diseases like cancer, heart disease, and immune conditions.
It functions similarly to large language models (LLMs) like ChatGPT but is adapted to handle medical data, where each diagnosis or factor is treated as a "token" to learn the "grammar" of disease progression.
The model has been validated across different populations, showing it can generalize from UK data to Danish health records without retraining, and its predictions align closely with observed disease trends.
Delphi-2M's generative nature allows it to create synthetic health trajectories, which can be used to train other AI models while preserving patient privacy.
Experts highlight its potential for personalized medicine and preventive healthcare, though they caution it is not yet ready for clinical use due to potential biases in training data related to age, ethnicity, and healthcare access.
The model is currently a research breakthrough, with plans to test its adaptability in more diverse populations and healthcare systems.
Posts
1 comment:
I'll take a real live Human doctor thanks very much.
Post a Comment