It's called S-Race, at San Raffaele the AI ​​on clinical data is already a reality

On the same topic:

At the San Raffaele Hospital – Gruppo San Donato – a team of researchers and scientists led by Carlo Tacchetti and Antonio Esposito, coordinators of the strategic AI program at the Vita-Salute San Raffaele University, has managed to “bend” Artificial Intelligence to the most important need of the human race: health . In tandem with Microsoft, the working group has developed a series of models – the result of an analysis produced by the AI ​​capable of “photographing” in detail the health conditions of patients with the same pathologies. “ It is called S-Race – confirms Professor Esposito – and it is an advanced digital platform, designed to collect, classify, standardize and analyze a large amount of real world health data from the hospital, ensuring full compliance with privacy regulations . The main objective is to develop predictive models for prognosis and response to specific therapies for each individual. These models will support the experience of doctors, allowing unprecedented personalization of the diagnosis and treatment of numerous pathologies. Currently, the platform already integrates 21 projects dedicated to different clinical areas including oncology, cardiovascular and metabolic diseases, intensive care and neuroscience. The diagnostic imaging sector will also benefit from this technology, optimizing the analysis and interpretation of data”.

“The idea for S-Race,” Professor Tacchetti explains to Il Foglio, “was born during the difficult years of the COVID-19 pandemic. In 2020, faced with the enormous influx of patients, San Raffaele asked Microsoft to collaborate to create a predictive model. This model, in just 30 minutes from the patient's arrival at the reception, allowed us to know if he had to be hospitalized due to the severity of his condition or could return home in a short time. Microsoft made available a group of technicians in the US, one in Europe and one in Italy, we deployed our team, with all the available data and in a short time we defined a model that with 5 predictive parameters allowed the doctor to define whether the patient was high risk or not”. During this process, the team developed standardized protocols and procedures in line with the regulations and recommendations for responsible use of AI. This is how the first prototype of S-Race was born, a secure, certified platform capable of developing models, following the principles of responsible and explainable AI. Operational since June 2024, S-Race is unique in the world in its ability to extract data from the main hospital databases, including medical records, classify them according to international medical ontologies and use them to develop both “black box” and “white box” models (we'll talk about this later).

Let's try to focus on what led the researchers at San Raffaele to this choice, also to understand who is in charge: medicine or artificial intelligence? "In engineering, two plus two always makes four - continues Tacchetti - but in medicine this is not true, there is a delta of uncertainty due to the partial knowledge of the human body and the existence of unknown variables. This makes each patient unique, even with apparently identical pathologies. The current challenge is to identify the variables that make patient A different from patient B, to provide a more accurate diagnosis, prognosis and therapy ". The enormous amount of health data generated every year (with a 36 percent growth worldwide) makes AI an indispensable tool for analysis. Not only that, what is the point: the time available to doctors. An oncologist has 20 minutes to perform a visit (regional agreement), give the therapy and write the medical record, so I need help and to analyze all the available data I need artificial intelligence. “And here comes S-Race, which addresses this challenge using two main approaches for the development of artificial intelligence models: the 'black box' model, where I give the computer a series of data and it provides me with a solution without explaining the reasoning followed and therefore I do not know why one patient goes to the right and one goes to the left. This approach, although effective, has a lower 'explainability', which is crucial in medicine because for reasons of liability the doctor must explain to the patient the proposed therapy” .

And the “white box” model? “Instead it is similar to a decision tree, which allows me to follow the logical path of the algorithm, offering greater transparency and understanding. Here at San Raffaele, standard procedures have been defined that combine both approaches, integrating clinical questions with the capabilities of AI” – concludes Tacchetti. Thanks to complementary funding to the Pnrr (D34Health - Digital driven diagnostics, prognostics and therapeutics for sustainable health care), under the appropriate conditions, the S-Race platform is exportable and will soon be available to other centers through a “federated” system . This indicates a future in which AI will be an increasingly valuable ally for medicine, supporting doctors in making informed and personalized decisions, but without replacing human experience and judgment. To conclude we can say: medicine beats AI, two to one .

More on these topics: