DD-CovidNet Model for X-Ray Images Recognition of Coronavirus Disease 2019

Abstract

Affected by the shortage of medical resources and low level of medical care, coronavirus disease 2019(COVID-19) has not yet been contained. It is a safe and effective way to detect infection in chest X-ray (CXR) images by deep learning. To solve the above problems, an intelligent method for automatic recognition of COVID-19 in CXR images is proposed. According to the characteristics of CXR images, a dual-path multi-scale feature fusion (DMFF) module and dense dilated depthwise separable (D3S) module are proposed to extract the shallow and deep features respectively. On this basis, an efficient and lightweight convolutional neural net-work-DD-CovidNet, is designed. DMFF module can sense more abundant spatial information by fusing multi-scale features. D3S module can extract more effective classification information by enhancing feature transfer and enlarging receptive field. The method is validated on two data sets. The experimental results show that the sensitivity of DD-CovidNet model for COVID-19 recognition is 96.08%, the precision and specificity are 100.00%, and it has less parameters and faster classification speed. Compared with other models, DD-CovidNet model has faster detection speed and more accurate detection results. © 2021, Beijing China Science Journal Publishing Co. Ltd. All right reserved.