Predicting the Severity Of patients with Coronavirus Using neuronal networks

Abstract

The new coronavirus (COVID-19) spread in 2019, starting from Wuhan in China, then spread quickly in many countries around the world, and infections exceeded (10 million) cases worldwide, according to statistics of The World Health Organization (WHO) till to 23 June 2020.

Victims of coronavirus attend hospitals daily in increasing numbers due to the rapid transmission of infection , which causes pressure on these hospitals and their resources, so the treatment provided depends on the degree of severity of the infected case, Delayed treatment of critically patients increases the number of deaths.

This research aims to build a classifier prediction model to predict the severity of patients with Coronavirus to give them priority in treatment and care, and to support the medical decision in managing them as quickly as possible.

The proposed model predicts mortality risks depending on different variables: demographic data, physiological symptoms of patients, the radiological findings of the chest x-ray , computed tomography CT scan and laboratory findings of daily blood tests, which are one of the least costly, effort and time diagnostic methods.

The results showed the efficiency of the proposed classification model in predicting the cases of recovery and death with the highest classification performance with accuracy of 95.9%, in addition the factors (cough, fever and the presence of interstitial opacities in chest radiographs and low oxygen saturation) increase the probability of death of infected patients and thus we can control the virus during them, so early detection of infection with the virus helps to manage it in time.