Diagnosis of Covid -19 Disease from Lung Images with Deep Learning


Department of Computer Science Faculty of Computer Uuniversity of Tabriz,Iran



In this research, an intelligent CAD system for covid disease is presented based on lung images. This system is based on deep learning. For this purpose, the design and construction of the CAD system is first performed for each desired image, pre-processing operations are performed in order to improve the quality and contrast of the edges of the image, to eliminate the noises on the image and to remove the redundant parts of the image. Then, using the deep learning technique, the segmentationoftheinputimageandtheextraction of important parts are done. After extracting the main parts, using deep learning architectures such as convolutional neural network, as well as combined transfer learning architectures, the feature extraction process is carried out in order to extract the features of segmented images. In the last step, because the features are available, the classification process is performed based on the features, and therefore the speed of the system increases. Next, a neural network architecture based on deep learning and the combination of different architectures is designed to classify and identify the type of input image between healthy and Covid-19 disease groups, and the process of training, testing and validation is implemented on it. It is possible to use the desired model to identify the covid disease. The evaluation of the results shows that the final accuracy obtained for the classification of the first scenario (healthy and viral) and the second scenario (healthy, covid, pneumonia) is about 99 and 95%, respectively. In addition, the results of the proposed method are promising in terms of accuracy, sensitivity and specificity criteria compared to recent researches; So that for the classification of the first scenario, it has a sensitivity and specificity of 100 and 99%, respectively. The proposed method can be used as a doctor’s assistant during the treatment of patients by finding a way to the application field.
Keywords: covid disease - prediction of covid disease - deep learning