Comparative Analysis of Deep Learning Models for Pneumonia Detection in Chest X-Ray Images

Abstract

This paper focused on Comparative Analysis of Deep Learning Models for Pneumonia Detection in Chest X-ray Image. Pneumonia is one of the illnesses which are associated with the lung’s region, which can lead to a severe condition when not diagnose or detected at early stages. The ability of the disease to restrict the flow of oxygen getting into the bloodstream makes the disease more dangerous as a result of existence of virus, bacteria or Fungi in the lung. Hence leads to untimely death. Experimental AlexNet ANN, ResNet50 ANN and DenseNet 121 ANN algorithms were to distinguish and detect pneumonia from non-pneumonia patients using medical images with AlexNet with a total number of 1877 images for both pneumonia and non- pneumonia patients were used to train the alexnet algorithm and 805 images of both pneumonia and non- pneumonia images were used for testing, the dataset contained a balanced combination of both pneumonia images and non-pneumonia images. The following results were gotten from the experiments for both AlexNet ANN and ResNet50 ANN respectively: the accuracy was 0.877, Sensitivity 0.834, specificity 0.917, f1Score 0.866 and the AUC which was 0.93; 0.817, Sensitivity 0.720, specificity 0.910, f1Score 0.793 and the AUC which was 0.88 and 0.915, Sensitivity 0.837, specificity 0.990, f1Score 0.906 and the AUC which was 0.98 with the Accuracy, Sensitivity, specificity and AUC values. The three Scenarios on three ANN Architecture were observed. It was found that all the three models were able to distinguish and detect pneumonia accurately with no significant error.