Enhance the Accuracy of Pneumonia Detection Using Feature Fusion and Artificial Intelligence

Abstract

Abstract— Pneumonia is a deadly bacterial infection that affects the lungs in the human body. According to the World Health Organization (WHO), pneumonia causes one out of three mortalities in India. Chest X-ray (CXR) images are utilized to diagnose pneumonia and it must be examined by a professional radiotherapist. Thus, developing an autonomous system for identifying pneumonia would be beneficial for prompt treatment. In this research, we used Artificial Intelligence (AI) and feature fusion methods to identify pneumonia from CXR images. The images are collected from Kaggle and processed for resizing, noise removal, and data balancing. The processed images are given to Deep Learning (DL) models like Inception and Xception to extract the features. Then, Canonical Correlation Analysis (CCA) is employed to fuse the features obtained from both DL models. The Inception, Xception, and fused features are given to Machine Learning (ML) models such as Support Vector Machine (SVM) and Bayesian Logistic Regression (BLR) for pneumonia classification. Finally, each feature with the ML model is evaluated individually using metrics such as accuracy, specificity, recall, precision, and F1 score. The experimental outcome shows that the CCA fused features with BLR outperform the other combinations, achieving the highest accuracy of 98.91%.