In a new work, a research group led by Hiroyuki Yoshida found that if an AI learns uncontrollably from the results of computed tomography (CT), it will be able to predict with a high degree of probability how a patient will have a disease.
The model determines, for each patient, how COVID-19 will progress, as well as the time to enter intensive care. In addition, with the help of AI, patients can be divided into groups of low and high risk of severe disease.
Our results show that predictions from the unsupervised AI model were made with greater accuracy than other similar designs. Now you can make predictions about the development of the disease immediately based on CT data.
Early versions of AI predicting the outcome of the disease were limited by various subjective assessments, semi-automatic prediction schemes, or other control methods. In this case, the percentage of forecasts that came true declined.
The authors of the new work believe that their development can be effectively used not only for a new type of coronavirus, but also for other serious diseases: you only need to adapt the data, and the basis is already there.