Early Detection of Covid-19 Disease using Computed Tomography Images and Optimized CNN-LSTM

Abstract

Since the novel Coronavirus (COVID-19) pandemic showed up in China, it became a big problem for health authorities to counter this life-threatening disease. Early light signs such as fever and nonproductive cough give a chance for early detection of disease and appropriate treatment. Imaging features that can be obtained using computed tomography (CT) images are of the most significant aspects of COVID-19 for screening, examination, therapy evaluation, and follow-up. This paper proposes an intelligent method for early detection of COVID-19 based on CT images and deep neural networks. In the developed method, the convolutional neural network (CNN) is used for automatic feature extraction from CT images and long-short term memory (LSTM) is used for final classification. Moreover, the Harris hawk optimization (HHO) algorithm is implemented for finding the best possible value of internal parameters of CNN and LSTM, such as the number of convolution/pooling layers, size, and the number of convolution kernels with the aim of increasing the classification accuracy. The developed method tested on data collected in Mashi Daneshvari Hospital in Iran. The obtained results showed that the developed method could detect the COVID-19 with high accuracy without needing radiologist experts.