A revolutionary AI tool has been developed for doctors to forecast the onset of serious health conditions like cancer and heart attacks. The AI program has been trained to predict over 1,000 diseases and could potentially assist healthcare professionals within the next five to ten years. By analyzing anonymized patient records, researchers have fine-tuned the model to anticipate individuals’ health outcomes over an extended period.
Ewan Birney, a key figure in the research from the European Molecular Biology Laboratory (EMBL), envisions a future where clinicians will be empowered by advanced AI tools. These tools will enable doctors to identify significant health risks for patients and suggest actionable steps to mitigate them. Common advice like weight loss and smoking cessation will likely remain standard, but tailored recommendations for specific diseases could become more prevalent.
The AI tool was trained using data from 400,000 individuals in the UK Biobank and validated with information from 1.9 million patients in the Danish National Patient Registry. It functions by assessing the likelihood of developing various diseases, such as cancer, diabetes, cardiovascular issues, and respiratory conditions, akin to predicting weather probabilities.
Notably, the tool excelled in predicting diseases with predictable progression patterns, like certain cancers and heart attacks, but showed less accuracy for conditions with fluctuating courses, such as mental health disorders or pregnancy-related complications. Moritz Gerstung, a leader in AI oncology at the German Cancer Research Centre, believes this technology marks the beginning of a new era in healthcare, offering personalized care and early intervention on a large scale.
The AI model considered medical history, lifestyle factors (like obesity and smoking), age, and gender to project disease risks based on the sequence and timing of events. Published in Nature, the study revealed that the AI tool, named Delphi-2M, could forecast disease rates for a wide range of conditions with accuracy comparable to traditional single-disease models. Its generative approach also allows for simulating future health trajectories, providing valuable insights into potential disease burdens up to two decades ahead.