Towards Automatic Diagnosis of the COVID-19 Based on Machine Learning

Abstract

Since the beginning of the global health crisis attributed to the new coronavirus (COVID-19), several announcements of new diagnostic tests have been made. It has suddenly become complicated to provide a comprehensive overview detailing the specificity of each of them. In the absence of therapeutic drugs or vaccines specific for COVID-19, these tests are essential to detect patients at an early stage and to immediately isolate infected patients from the healthy population. Among the most commonly used tests are chest CT and laboratory tests (such as PCR) in the diagnosis of coronavirus 2019 (COVID-19). Analysis of imaging and laboratory test data from more than 1,000 patients shows that chest CT surpasses biological tests in the diagnosis of the epidemic associated with the new coronavirus, COVID-19. Thoracic computed tomography (CT) scans provide the best diagnosis for COVID-19 pneumonia, conclude these researchers from Huazhong University of Science (Wuhan, China) and Leiden University Medical Center (Netherlands). The researchers concluded that CT scans should be used as the primary screening tool for COVID-19. However, these techniques have the disadvantage of being slow and expensive, causing many patients to avoid screening if it is not free. In this paper, we propose an automatic decision model based on artificial intelligence to assist the physician during the screening for this pandemic in order to lighten the workload in hospitals. For this purpose, the main objective of our study is to build and test predictive diagnostic models based on machine learning that can analyze patient data in depth in order to separate coronavirus types. Thus, we can save time during the diagnostic process of this Covid-19 pandemic. The results obtained are very promising.