For this yes. This is one of the clearest examples of positive use of artificial intelligence. A team of scientists fromEuropean Molecular Biology Laboratory (EMBL) has developed a generative AI model that uses large -scale medical records to estimate how human health can change over time. You can forecast the risk and the moment of appearance of more than 1000 diseasesas well as predict health results with more than a decade in advance. The advance has been published in Nature.
This new generative the AI model was designed using algorithmic concepts similar to those used in large -scale language models (LLM). He trained with Anonymized data of patients of 400,000 participants of the United Kingdom biobanco.
So that the algorithms did not fall into errors caused to obtain the information from a single data source, lThe study authors also successfully tested the model using data of 1.9 million patients of the National Registry of Patients of Denmark. This, according to the team led by Ewan Birney, is one of the most complete demonstrations to the date of how generative AI can model the progression of human diseases on scale and was tested with data from two completely separate health systems.
“Our AI model is a proof of concept that shows that it is possible for AI to learn many of our long -term health patterns and use this information to generate significant predictions – explains Birney, In a statement -. When modeling the evolution of diseases over time, we can begin to explore when certain risks and The best way to plan early interventions. It is a great step towards more personalized and preventive approaches to medical care.
Like the great linguistic models, they can learn the structure of the sentences, this model of AI learns the “grammar of the health data”. This allows you to model medical records as sequences of events that are developed over time. How to establish the steps on a map that take us from A to B. Only with thousands of data. These events include Medical diagnoses or lifestyle factors such as smoking. The model learns to predict the risk of disease from the order in which these events occur and of the time elapsed between them.
“Medical events usually follow predictable patterns – adds Tom Fitzgerald, co -author of the study – our AI model learns those patterns and can forecast future health results. It allows us to explore what could happen based on the medical history of a person and other key factors. Fundamentally, this is not a certainty, but an estimate of potential risks. ”
The model works especially well in conditions with clear and consistent progression patterns, such as certain types of cancer, heart attacks and septicemia. However, it is less reliable for more variable conditions, such as Mental health disorders or pregnancy -related complications that depend on unpredictable vital events.
Despite being a great step, huge in terms of prevention, several aspects must be highlighted. The model, which is not yet ready for clinical use, provides probabilities, not certainty. It does not predict exactly what will happen to a person, rather offers well -calibrated estimates of the probability that certain conditions are presented in a certain period. For example, the probability of developing heart disease could predict during the next year.
You also have to keep in mind that The data were obtained from volunteers between 40 and 60 years, thus children’s and adolescent health events are sub -present. The model also presents demographic biases due to the lack of training data, including the low representation of certain ethnic groups.
That said, it does allow a pattern of possible patients of many types of conditions taking into account Many of its customs and create prevention or early detection measures of tumors, for example.
The idea is that, in the future, similar to tools, trained with more representative data sets, could help doctors and scientists to identify high -risk patients. Something that, seeing The aging of the population and the increase in chronic disease rates sounds like very good news.
“This is the beginning of a new way of understanding human health and the progression of diseases -concludes Moritz Gerstung, co -author of the study -generative models such as ours They could someday help to customize medical care and anticipate large -scale needs. ”