Automatic Diagnosis of Pneumonia and COVID-19 Using Convolutional Neural Networks and Transfer Learning

Abstract

Several studies are currently exploring the diagnosis of lung disorders using deep learning analysis of medical images. Deep learning is also considered to be a valuable aid to experts in the interpretation of medical images. Heuristics such as transfer learning are becoming more common; these methods (based on pretrained models) are utilized as the basis for computer vision tasks and can significantly improve various issues. This work proposes models built on Convolutional Neural Networks (CNNs) that incorporate transfer learning to identify various pneumonia infections in X-ray images. The experiments show that the model based on Xception network outperforms many existing state-ofthe- art methods and several recent backbones.