Comparative Analysis of Pneumonia Detection using Deep Neural Networks on Chest X-ray Images

Abstract

Pneumonia is a lung infection that causes inflammation in the alveoli and remains a major health concern not only in India but also across the world due to its high mortality rate. Novel therapeutic strategies are required to address pneumonia, a complicated pulmonary disease. A number of medical diagnostic applications have recently seen a surge in the use of deep learning (DL) techniques. Examining how well DL models identify and categorize pneumonia from chest X-ray (CXR) pictures was the primary goal of this study. This study focuses on improving the performance of automatic pneumonia detection from chest X-ray images by optimizing model parameters in deep learning architectures. In this research, pre-trained convolution neural network (CNN) models, Resnet-50, VGG-19, EfficientNetB0, and a proposed DEEP neural network (DNN) models have been developed. The experimental results show that two models, one pre-trained (EfficientNetB0) and proposed DNN achieve strong classification performance, with accuracy 89.2% and 90% respectively.  Proposed DNN model consistently delivers the highest accuracy and very good statistical parameter performance with KAPPA 82%, F1-score 94.0% and AUC 96.5%. These findings indicate that proposed DNN model is moderately an effective balance between accuracy and other parameters.