A Lightweight Depthwise Separable Convolution Neural Network for Screening Covid-19 Infection from Chest CT and X-ray Images

Abstract

Because Covid-19 spreads swiftly in the community, an automatic detection system is required to prevent Covid-19 from spreading among humans as a rapid diagnostic tool. In this paper, we propose to employ Convolution Neural Networks to detect coronavirus-infected patients using Computed Tomography and X-ray images. In addition, we look into the transfer learning of a deep CNN model, DenseNet201 for detecting infection from CT and X-ray scans. Grid Search optimization is utilized to select ideal values for hyper-parameters, while image augmentation is employed to increase the model’s capacity to generalize. We further modify DenseNet architecture to incorporate a depthwise separable convolution for detecting coronavirus-infected patients utilizing CT and X-ray images. Interestingly, all of the proposed models scored greater than 94% accuracy, which is equivalent to or higher than the accuracy of earlier deep learning models. Further, we demonstrate that depthwise separable convolution reduces the training time and computation complexity.