Deep learning model for predicting COVID-۱۹ in x-ray images

Background and Aim : COVID pandemic has been one of the most severe and life-threatening phenomena, causing many deaths worldwide. Despite the discovery of various vaccines, nations have not yet been able to control COVID. Therefore, rapid detection at its early onset is one of the most crucial steps toward preventing its prevalence. Lung is the first endangered organ during this infection. Therefore, urgent diagnosis and treatment of this organ are of the main concerns currently. Chest radiography is one of the practical steps to achieve this purpose. The only challenge in diagnosing CXR-based COVID-۱۹ patients is that trained physicians may not always be available, especially in remote areas. Moreover, the radiological manifestations of COVID-۱۹ are new and unfamiliar to many professionals who have no previous experience with COVID-۱۹-positive CXRs. Methods : In this paper, we introduce one of the newest methods in deep learning science using advanced models in data mining and the detection of complex patterns to detect damaged lungs. A database including ۲۰۰۰ images of healthy lungs and ۸۴ images of patients was used to train the system. However, due to the lack of data for the training system, the data volume was increased using data generation methods. After that, a dataset including ۲۰۰۰ images of healthy lungs and ۱۰۰ images of patients was used to test the model. It should be noted that to increase the accuracy of our model, in addition to healthy people, patients with other kinds of diseases have been added to this database. Results : Using the DenseNet structure, a model with an accuracy of ۹۸.۵% in diagnosing this disease is presented in this study.Conclusion : The results show that this model is accurate enough in diagnosing Covid disease along with other diseases. The program can be easily used on any computer and mobile phone by any medical staff to diagnose COVID patients using chest X-ray images in seconds.