This tool could predict your risk of cancer or heart disease years in advance. Here’s how
The AI model that could predict illness decades before symptoms appear
- European researchers have developed Delphi-2M, an AI tool designed to predict a person's next health event and its likely timing over the next 20 years for a thousand different diseases.
- The model utilises a 'transformer network', similar to ChatGPT, and was trained using data from nearly 403,000 individuals from the UK Biobank.
- Delphi-2M predicts health outcomes based on factors like sex, BMI, lifestyle, and prior diseases, achieving a theoretical accuracy of approximately 70 per cent in initial tests.
- Unlike previous models, Delphi-2M can predict multiple diseases and is open-source, allowing other researchers to adapt and improve it using synthetic data without compromising privacy.
- While not yet ready for clinical use due to data diversity and scalability concerns, the model demonstrates the potential for future AI tools to evolve with richer data for personalised health predictions.