A CNN based Hybrid Model for Pneumonia Classification Using Chest X-ray Images

Abstract

Pneumonia is a lung infection caused by bacteria, viruses, or fungi. It is one of the deadliest lung diseases among children under the age of five. An expert or radiologist can usually diagnose the condition using X-ray images of the chest. The use of machine learning in medical image processing helps to improve detection accuracy. This study aims to develop and present a combined Deep Learning model for classifying patients with Pneumonia disease based on chest X-rays. Three separate models are trained for the chest X-ray dataset in the proposed implementation, the first of which is the custom Convolutional Neural Network model. The other two models are Xception and EfficientNetB4. Various data augmentation and pre-processing methods are used, along with hyperparameter tuning. A combined model is created by assigning weights to the trained models based on their recall and accuracy values, and the classification results are obtained by a polling mechanism at the output, which gives an accuracy of 98.00%. The proposed work outperforms the existing literature in terms of several performance parameters.