Researchers Develop AI System That Predicts Over 1000 Diseases Years Before Onset
In a scientific development considered a breakthrough in the field of preventive medicine, an international research team announced the development of a new AI model capable of predicting the likelihood of a person developing future diseases years before they occur.
The model, named (Delphi-2M), relies on techniques similar to those used in chatbots, as it learns patterns and sequences in medical data. According to the researchers, the tool is capable of predicting over 1000 diseases, opening new horizons for early intervention and prevention of diseases.
The model was trained on the vast data recorded in the UK Biobank, which includes vital medical information for approximately half a million participants. The system analyzes the complete medical history of the patient, previous diagnoses, and the sequence of their occurrence.
In this regard, Moritz Gerstung, an AI expert at the German Cancer Research Center, explained: "Understanding a series of medical diagnoses is somewhat similar to learning the rules of grammar in a text." He added that the tool (Delphi-2M) "learns patterns in healthcare data, previous diagnoses, the combinations in which they occur, and the sequence, enabling highly meaningful and health-relevant predictions."
To test the model's accuracy, the team experimented with data from nearly two million people in Denmark's public health database. The results showed the system's ability to identify individuals most at risk of diseases such as heart attacks, surpassing predictions based on age and other traditional factors.
However, the researchers warned that the tool "is not ready for clinical use yet" and requires further testing. Peter Bannister, a health technology researcher and fellow at the British Institute of Engineering and Technology, pointed to the challenge of data bias, stating: "Researchers acknowledge that both datasets (British and Danish) are biased in terms of age, race, and current healthcare outcomes."
Despite the challenges, scientists see enormous potential in this technology. Tom Fitzgerald, a co-author at the European Molecular Biology Laboratory, stated that these tools could help "improve resources across an overstretched healthcare system."
One unique advantage noted by Ewan Birney, a co-author of the research, is that the tool (Delphi-2M) "can diagnose all diseases at once over a long period," unlike current tools that focus on specific diseases.
Gustavo Sudre, a professor at King's College London and a specialist in medical AI, concluded that the research "seems to be an important step towards scalable and interpretable predictive modeling, and most importantly, ethically responsible."
It is noted that doctors in many countries are already using predictive tools such as the (QRISK3) program, but the new system represents a qualitative leap in terms of the range of diseases and the time frame it predicts.
