An Efficient Approach for Automatic detection of COVID-19 using Transfer Learning from Chest X-Ray Images

Abstract

The coronavirus disease 2019 (covid 19), which was declared a pandemic by the World Health Organization (WHO) in December, causes significant alveolar damage and progressive respiratory failure, resulting in death. The only laboratory technique available, RT–PCR, has an accuracy of about 73 percent. Medical specialists may benefit from early detection using CXR. Using deep convolutional neural network architecture, we propose a Com-puter Aided Diagnosis (CADx) for the diagnosis of coronavirus disease 2019.The chest x-ray dataset is used for testing and training of neural networks. The CXR images are segmented using a U net model, and the segmented image is then used to train a classification model using the Inception v3 model, which distinguishes covid 19 from pneumococcal records and safe records. Training of inception v3 is done with different resolutions of Chest X-rays (CXR) and for further optimization adam optimizer is used. This model produces high computational efficiency with an accuracy of 0.97 per-cent. Based on the promising results obtained the proposed method can be used for effective diagnosis of covid 19 during this pandemic.