The new coronavirus is highly contagious. However, despite the wide spread of the virus and its ability to bring entire healthcare systems to their knees, only a small percentage of patients - around 5-10% - develop the most serious (and sometimes fatal) forms of the disease. Often these forms have a rapid and unpredictable course that lead the patient to go from a mild symptomatology to a serious respiratory insufficiency in a very short time. The San Raffaele Hospital in Milan, part of Euro-BioImaging’s Molecular Imaging Italian Node, is contributing to an important and timely project called AI-SCORE, along with Microsoft and NVIDIA, to calculate prognostic risk from COVID-19, based on CT imaging and artificial intelligence. Carlo Tacchetti, Project coordinator, and Antonio Esposito, namely director and deputy director of the Center of Experimental Imaging of the IRCCS San Raffaele Hospital in Milan, explain how CT imaging is an important component of this study.
The aim of the project is twofold: on the one hand to recognize in the general population the people at greatest risk of developing severe forms of Covid-19, if infected with the virus, that should be protected the most; on the other hand, to recognize among the patients who show the first symptoms of Covid-19 those who will have the worst prognosis. The project will start from this second objective, with an AI algorithm that will integrate diagnostic images (images of the ungated chest CT), clinical and laboratory parameters, inflammatory status, and genetic profile of the patient and the virus.
The technological platform is based on the latest Microsoft innovations for data analysis and Artificial Intelligence and will allow the collection, processing, management and use of heterogeneous data, coming from multiple sources, in total respect for privacy, to provide medical and research personnel timely and detailed information useful to support the decision-making phase and the processes necessary to respond to the various phases of the emergency. Microsoft also makes available a large group of partners that on its platforms develop solutions capable of re-inventing healthcare.
“The major challenge of this project is the use of Artificial Intelligence to drive a multidimensional and multi-parametric approach to associate chest CT imaging features with over 100 genomic and clinical parameters,” explains Dr. Carlo Tacchetti. “to identify factors relevant to define a prognostic risk score for COVID-19 patterns at the onset of the initial symptoms of the disease. To this aim, we will use images of the ungated chest CT, obtained as part of the clinical assessment of the lung and analyzed to measure quantitative markers of lung and cardiovascular injury.”
The project will be developed in three main phases: a first phase of data collection and homogenization of over 2000 patients hospitalized in the past weeks and with a known prognosis; a second phase of development and implementation of the algorithm, which will be "trained" to learn how to combine them in an "intelligent" way to predict the risk of the individual patient, and a third phase of testing and validation of the product on a second cohort of patients and in any prospective studies.