A Review of Approaches and Difficulties in the Use of Deep Learning Techniques for COVID-19 Disease Diagnosis on Chest X-Rays

Abstract

The global community's healthcare system has faced significant challenges as a result of the COVID-19 outbreak, which has increased demand for prompt and accurate diagnosis. Of these techniques, deep learning-based (DL-based) models have shown a great deal of promise in identifying COVID-19 from medical imaging techniques, especially CT and chest X-ray scans. The latest developments in COVID-19 detection using chest X-ray images are presented in this survey paper. As discussed, various deep learning architectures, including CNN, ResNet, Xception, and hybrid models, have shown promise in accurately classifying cases of COVID-19, pneumonia, and normal lung disease. Furthermore, a comparison is made between the accuracy, sensitivity, specificity, and F1-score level of each individual ADA and ADA with each model. Along with highlighting current drawbacks like data imbalance and overfitting, the paper suggests future development directions to give the model resilience and a range of applications.