Experimentation – It all started a year ago when Professor Bianchi and a team of colleagues present a study to use artificial intelligence to study images in the medical field: the so-called diagnostic imaging. The study, in collaboration with the nuclear medicine department of the Galliera hospital in Genoa, was presented to recognize the symptoms of Parkinson’s disease but the coronavirus emergency has led to further development. “For a long time I have been very close to the doctors of the Bergamo hospital – continues Bianchi – and since the end of February we have been in constant contact, but only from the human point of view and not to talk about medicine: they told me that at the Pope Giovanni hospital at the beginning March had exaggerated access, even 60-70 per night. At the beginning, to recognize the Covid 19 cases, the lungs were made the most accurate tool, but since the cases have become thousands, lung tests were often done with the chest plate. ”
Doctor Bianchi understands that with so many infections, a way to recognize the patients will be needed faster and so, after consulting with the nuclear physicist of the University of Pisa Roberto Cappuccio, presents the study to the Radiology department of Bergamo directed by Dr. Sandro Sironi that accepts the request. They came from Pope John XXIII two hundred chest plates while another hundred from the hospital in Massa Carrara: it is on these that the Nuclear Medicine team, in collaboration with the hospital’s Radiology see Massa Carrara and the Nuclear Physics Institute of the University of Pisa, is working to make the instrument usable within a few months.
How does it work – The two ongoing studies, coordinated by Professor Cappuccio, aim to study the spread of the virus through artificial intelligence. The first, through the analysis of the CT for pneumonia from Covid 19 from 1 November 2019 to 31 January 2020 in Tuscany and Bergamo, will serve to understand if and how much the virus circulated in Italy even before the official announcements: what, in science, one is called retrospective study. The second, however, linked to the analysis of the thoracic plates, will serve above all for the future and to trace the virus in the coming months. “We hope that a second wave does not come – explains the physicist Cappuccio – but in that case this tool will give us the opportunity, through Usca units, to identify covid pneumonia in seconds and directly to the patients’ home“.
One is used to do this RetinaNet neural network, a sort of software that can analyze images and bring out some specific characteristics to recognize the disease. This network will be “trained” by the same scholars – physicians and physicists – who will insert in the software the typical characteristics of viral Covid 19 pneumonia and with a series of distinctive features from other pneumonia. Once “trained”, the network will be able to recognize the symptoms on the body from thousands of plates in a few seconds with a 97% accuracy. The system can also be used by means of small and portable machines capable of carrying out x-rays even at the home of patients and not in the hospital. “This operational study may become concrete in the coming weeks – concludes physicist Cappuccio – could be decisive for the prevention and for a possible second wave of the virus”.