Analysing COVID-۱۹ in Medical Images

The Covid-۱۹ virus was discovered in Wuhan-China in ۲۰۱۹ and spread rapidly throughout the world dueto its high transmission power. A timely and accurate diagnosis of Covid-۱۹ is essential to the patient'srecovery. Based on deep learning algorithms and CT images, this study proposed hybrid methods todiagnose COVID-۱۹. First, we use wavelet transformation in combination with fuzzy logic to provide anew approach to removing the noise of CT images. Then we segmented lung images by the proposedcombined global and local threshold method. In this way, lung regions from CT images can be segmentedsuccessfully. In the next step, features and classification will be extracted. AlexNet is used to extractfeatures, while a Support Vector Machine (SVM) is used for classification. With ۹۹.۸% accuracy, threeclasses of data are classified: COVID-۱۹, Viral Pneumonia, and Normal. In comparison with previousmethods, the proposed method shows superior classification performance.

Keywords:

COVID-۱۹ , Lung Segmentation , Support vector machine (SVM) , AlexNet , Convolutional neural networks