Automated Pneumonia Classification from Chest X-Rays Using VGG19: A Deep Learning Approach

Pneumonia is still one of the major contributors to morbidity and mortality in many parts of the world. Therefore, early diagnosis is immensely relevant so that better prognostication can be achieved in the affected patients. This study researches the deep learning-based categorization of chest X-ray images for the presence of pneumonia using the VGG19 convolutional neural network architecture. Our developed model will be trained and tested on the available dataset that consists of 234 normal images and 390 images with pneumonia. Among them, the VGG19 model reached an overall accuracy of 89.10%, hence it was able to classify most cases correctly. The precision of this model was 86.43%, which means this model had a very good performance in correctly identifying the positive pneumonia cases. Moreover, the recall rate was 97.95%, indicating that the model could detect most pneumonia cases, which makes it highly sensitive. However, the specificity was comparatively lower at 74.36%, which indicates that the model generated some false positives in the identification of healthy individuals. The F1 score, balancing both precision and recall, stood at 91.83%, underlining the robust overall performance of the model. These results show that deep learning models such as VGG19 can be quite useful in pneumonia diagnosis, particularly when access to radiologists is at a minimum. Further improvements should be made to increase the specificity of this model and reduce the false positive cases. Further effort will be related to trying to improve the accuracy of the model by different techniques such as data augmentation and also by further tests in clinical settings to evaluate its practical application in real-world settings.