In this article, we present how to create a database of Covid-19 diseases at National Hospital of Tropical Diseases (called the CovidVN database) and then develop a learning neural network based on this database to diagnose this disease. The CovidVN Database is built based on the processing of real diagnostic test results of Covid-19 patients with a large number of samples and in accordance with the structure of the Israeli Health System Covid-19 disease database (called COVIDIsr Database). Then two MLP Artificial Neural Networks corresponding to these two databases will be developed using the Deep Learning Toolbox of MATLAB software; the results of training these networks and their accuracy are compared with each other to assess the relative quality of the CovidVN database. Other, the paper presents the method of assigning weight corresponding to pathological symptoms for network input parameters. The results showed that the weighting of input attributes corresponding to the pathological symptoms is significant.