Researchers have developed an artificial intelligence model that can predict potential health issues up to ten years in advance, based on patterns identified in individuals’ medical records. This technology, referred to as Delphi-2M, aims to assess the risk of over 1,000 diseases, functioning similarly to a weather forecast that indicates a chance of rain.
The goal of the AI model is to identify high-risk patients for early intervention, potentially preventing disease and enabling hospitals to forecast local health demands. Delphi-2M employs techniques akin to those used by popular AI chatbots, trained to recognize patterns in anonymized medical records rather than language. While it does not provide specific dates for health events, it estimates probabilities for various diseases, such as a 70% chance of developing a condition.
Initially, Delphi-2M was created using data from the UK Biobank, encompassing health records from over 400,000 individuals. Its predictive accuracy was further verified against medical records from an additional 1.9 million individuals in Denmark. Researchers report that the model shows strong performance, particularly for diseases with clear progression, like type 2 diabetes and heart attacks.
While the AI tool is not yet ready for clinical application, it is intended to function in a similar manner to existing risk assessments for medication, such as statins for heart disease. Future applications may enable personalized medical advice and more effective disease screening, with the potential to anticipate healthcare needs in specific regions.
The model’s research, recently published in Nature, highlights the need for further refinement and testing before clinical use, especially given the bias inherent in using predominantly older UK data. The study represents a collaboration between several institutions, including the European Molecular Biology Laboratory and the German Cancer Research Centre, with ongoing efforts to incorporate additional medical data into future iterations.
Source: https://www.bbc.com/news/articles/cx2pj502ev6o?at_medium=RSS&at_campaign=rss