Deep Learning Approach for Diagnosis of Pneumonia

In the pandemic of Coronavirus disease 2019 (Covid-19), lung diseases are the leading cause of death worldwide. Among these diseases, pneumonia is the largest infectious disease caused by bacteria, viruses (including coronavirus) that collapses alveoli present in the lungs. Most effective and economic modality for diagnosing it is Chest X-Ray (CXR) imaging. In this paper we have proposed a scratch Convolutional Neural Network (CNN) 13 layers model for detecting pneumonia from CXR images. We have evaluated proposed CNN-13 model on pneumonia and normal CXR image dataset which is freely available on website of Kaggle. After simulation, the proposed CNN-13 model attained results with training and testing accuracy of 98.51% and 96.7% respectively. Metrics log loss and Area Under the Curve (AUC) scored 0.0371 and 0.9984 on training set and 0.1285 and 0.9819 on testing set respectively.