ABSTRACT
Coronavirus disease (Covid-19) is an infectious respiratory disease caused by SARS-CoV-2. Among the symptoms, the respiratory system of the sufferer is affected. This respiratory condition suggests that the chest imaging plays a key role in the diagnosis of the disease. Several pre-trained deep learning models have been developed to detect Covid-19 through chest radiographs. These models provide high precision for binary detection, however, when combined with other diseases such as pneumonia that also affect the respiratory system and lungs, they offer poorer quality performance and reduce screening performance. In this study, we analyze some neural networks models for binary and multiclass classification of X-ray images in order to find out the best accuracy of classification. The models are based on deep learning methodology to learn and classify images. They are extracted from well-known repositories such as Kaggle. The conducted experiments compare their performance in several scenarios: a multiclass classification model versus the combination of several binary classification models. Two methods for combining the output of the binary models are proposed to achieve the best performance. The results show that the best results are obtained with a well-trained multiclass model. However, a preliminary screening can be obtained from the binary models without creating and training a more complex model. © 2023, The Author(s), under exclusive license to Springer Nature Switzerland AG.