AI-SCoRE: predicting the course of disease among technology, medicine and computer science
At the height of the SARS-CoV-2 pandemic, emergency rooms around the world were saturated with patients with symptoms attributable to Covid-19. Due to the difficulty of predicting who would develop a severe form of the disease, a large number of patients were being admitted, wasting resources and overcrowding.
Two Vita-Salute San Raffaele University professors, Carlo Tacchetti and Antonio Esposito, proposed their ideas to Microsoft, a global computer science giant, and its partner companies Porini and Orobix, asking for collaboration to develop their own idea and create an accurate and rapid system to predict the level of clinical risk and define the most appropriate course of care for patients. This project was named AI-SCoRE (Artificial Intelligence – SARS-CoV-2 Risk Evaluation).
How the AI-SCoRE platform works
The platform developed for AI-SCoRE collects, standardizes and analyses patients’ data in full respect of their privacy.
In the early stages of the pandemic, hundreds of parameters were reported in scientific literature, each of which was individually associated with a severe or mild form of the disease: however, there was a lack of studies that could integrate these parameters to provide a univocal answer. Based on these parameters, the algorithm has defined five as the most predictive of prognosis: age, sex and oxygen saturation, which are easily detectable on access to the emergency room, and two radiomic parameters that the AI automatically extracts from patients' chest CT scans.
With this system, in about 30 minutes, the platform provides a probability of development, for each patient, of a mild, medium or severe form of disease. This result is notified to healthcare professionals on a dedicated app and helps to establish the best treatment: home monitoring, hospitalization in the ward or in the intensive care unit.
The algorithm was trained on data from over 1500 patients recruited from 16 Italian hospitals during the first pandemic wave, and prospectively validated on patients admitted to San Raffaele in the second wave. The project promised to significantly reduce inappropriate hospitalizations, with a classification accuracy of approximately 100%.
A unique collaboration
The challenge generated at the time of Covid-19 soon evolved into a strategic partnership between Vita-Salute San Raffaele University, IRCCS San Raffaele Hospital (San Donato Group) and Microsoft: a unique experiment in Europe that combines the clinical knowledge of multidisciplinary teams with the latest digital innovations.
The San Raffaele/Microsoft collaboration has yielded concrete results. One example, trials of AI solutions are being implemented for the treatment of non-small cell lung cancer, thus supporting the selection of patients eligible for immunotherapy. Other trials based on the AI platform are also in place to support the treatment of other diseases, including kidney cancer, diabetes, and cardiovascular diseases.
The AI-SCoRE project was co-financed with the POR FESR 2014-2020 resources.