AI model that can 'see' your illness decades in advance announced

Developed based on the same technology as ChatGPT, the Delphi-2M AI model is said to be able to predict all diseases at the same time and over a long period of time.

Delphi-2M is capable of "predicting the incidence of more than 1,000 diseases" based on patient medical records - Illustration photo

A group of European scientists announced on September 17 that they had developed an artificial intelligence (AI) model that can predict a person's risk of disease many years in advance, based on technology similar to chatbots like ChatGPT.

In a study published in the journal Nature, a team of authors from the UK, Denmark, Germany and Switzerland said the AI ​​system, called Delphi-2M, is capable of "predicting the incidence of more than 1,000 diseases" based on patients' medical records.

The model was trained with data from UK Biobank - a large UK biomedical database with information on around half a million participants.

According to AI expert Moritz Gerstung of the German Cancer Research Center, reading and analyzing medical diagnosis strings is "like learning grammar in a text". Delphi-2M can recognize patterns in medical data, the combination and sequence of diagnoses, thereby making "meaningful and valuable predictions for health".

The team tested Delphi-2M's effectiveness on data from nearly 2 million people in a Danish public health database, and found that the system could accurately identify people whose risk of heart attack was higher or lower than would be expected based on age and other factors.

However, the authors note that the tool is not yet ready for clinical use and needs further testing. Researcher Peter Bannister of the UK's Institute of Engineering and Technology said that current data is biased by age, ethnicity and medical conditions.

In the future, systems like Delphi-2M could help guide surveillance, early intervention in preventive medicine, and contribute to optimizing resources in the health system, according to co-author Tom Fitzgerald of the European Molecular Biology Laboratory.

Unlike existing tools such as the QRISK3 program used by GPs in the UK to assess the risk of heart attack or stroke, Delphi-2M "can predict all diseases at once and over a long period of time," co-author Ewan Birney asserted.

Professor Gustavo Sudre of King's College London said the research was an important step towards "scalable, explainable and ethically responsible" predictive models - which are at the heart of the field of medical AI.

Back to topic

VNA

Source: https://tuoitre.vn/cong-bo-mo-hinh-ai-co-the-nhin-thay-benh-tat-cua-ban-truoc-ca-chuc-nam-20250918135918347.htm