Covid-19 Detection based on Transfer Learning & LSTM Network using X-ray Images

ABSTRACT

COVID-19 is a virus which leads to infections in the upper respiratory system and lungs. On the scale of the global pandemic, cases and deaths are rising daily. X-ray is one test that can give a better picture of the severity of COVID-19. To monitor various lung diseases, chest X-ray imaging is helpful. This paper proposed techniques, viz. deep feature extraction and pre-trained neural networks (CNN) to distinguish COVID-19 and normal (healthy) chest X-ray images. For deep feature extraction, the pre-trained deep CNN model VGG-16 was used. An LSTM model is introduced in this study. The dataset contains 180 X-ray images of COVID 19 and 200 healthy ones used in the experimental analysis. The performance measurement of the research was based on categorizing accuracy. Experimental activities show that deep learning demonstrates the potentiality in detecting COVID-19 based upon chest X-ray images as examined;the introduced model accomplishes an average accuracy of 97.37%. Other strategies like Resnet50 give 82% accuracy, Inception gives 96% accuracy, and Xception provides 92% accuracy. This has shown deep mechanisms that work well compared to local descriptions of the method of curing COVID-19 based upon the chest X-ray images. These findings allow us to conclude that this article's proposed procedure may help clinicians determine COVID-19-related diagnoses. © 2022 IEEE.