Pneumonia Image Classification: Deep Learning and Machine Learning Fusion

This paper introduces a comprehensive method for diagnosing pneumonia in chest X-ray (CXR) images. By employing deep learning models for feature extraction and machine learning methods for classification, we evaluated the effectiveness of feature extraction using ResNet50, VIT, and Inception v3 models, and compared the performance of random forest and logistic regression classifiers based on these models. A comprehensive assessment of accuracy, F1 score, and AUC values was conducted to determine the most effective model. This study addresses the challenge of manual pneumonia diagnosis and emphasizes the urgency of automated solutions. Ultimately, the combination of Inception v3 and logistic regression achieved higher accuracy (79.32%), a higher F1 score (0.6239), and a more ideal AUC value (0.94). These results enhance the clinical efficiency and accuracy of pneumonia detection, benefiting patient treatment outcomes.