Predictive Analysis Using AI Model "Delphi-2M": Assessing Health Risks and Diseases
In a groundbreaking development, a team of researchers from England, Denmark, and Heidelberg University have created an AI model named Delphi-2M. This innovative model, based on a Transformer architecture similar to that used in large language models, is designed to predict the risk and progression of over 1,000 diseases up to 20 years in advance.
Trained using clinical data from 400,000 patients from the UK Biobank, Delphi-2M takes into account various factors such as body mass index and consumption habits. The model has been shown to achieve an average C-index of approximately 0.85 over a 5-year period, providing reliable predictions for the risk of heart attacks, certain tumors, and mortality, among others.
However, the potential benefits of AI models like Delphi-2M should be weighed against the ethical implications and potential risks of misuse or unfair disadvantage. Prof. Robert Ranisch, a Junior Professor of Medical Ethics with a focus on digitalization at the University of Potsdam, has expressed concerns about false expectations among insurers or employers, particularly outside Germany, due to the illusion of precise predictability.
Ranisch also raises questions about the implications of healthy people being classified as soon-to-be sick and the protection of health information when a multitude of personal data becomes relevant for AI predictions. He emphasizes the importance of considering the ethical considerations in the development and use of AI models in healthcare.
PD Dr. Markus Herrmann, head of the AI Ethics department at the Institute for Medical and Data Ethics at Heidelberg University, explains the distinction between using AI technology to assess developments in the entire healthcare system and making statements about individuals. Herrmann states that people have a right "not to know" - that is, a right "not to live their life in worry or even fear of impending illness."
The concerns raised by Ranisch and Herrmann highlight the need for careful consideration when it comes to the use of AI technology in making individual predictions. The use of AI models in healthcare could lead to people being unfairly disadvantaged if predictions are not reliable. As such, it is crucial to strike a balance between the potential benefits and the ethical implications of these advanced AI models.
The results of Delphi-2M have been published in the journal Nature, and initial tests show that the system provides similarly reliable predictions for the risk of various diseases as specialized models. However, the debate about the ethical implications of such AI models in healthcare is far from over, and it is essential to continue the discussion to ensure the responsible and ethical use of these technologies.
