ENHANCED PNEUMONIA DETECTION IN CHEST X-RAYS USING ATTENTION AND FNMS

Abstract

Pneumonia is a life-threatening respiratory infection 
that requires rapid and accurate diagnosis for effective treatment. 
In this study, we develop a deep learning-based pneumonia 
detection and classification model using chest X-ray images, 
distinguishing between normal, bacterial pneumonia, and viral 
pneumonia cases. The dataset, sourced from publicly available 
medical image repositories, is preprocessed and augmented to 
improve generalization. A Convolutional Neural Network (CNN) 
model is trained using optimized hyperparameters, with tech- 
niques such as batch normalization, dropout regularization, and 
early stopping to enhance accuracy and prevent overfitting. 
The model is evaluated on a separate test set, achieving a 
promising accuracy in detecting pneumonia subtypes. Further, 
performance metrics such as precision, recall, F1-score, and 
confusion matrices are analyzed. This research demonstrates the 
potential of deep learning in medical image analysis, offering 
a scalable and automated approach to assist radiologists in 
early pneumonia diagnosis. Future work includes leveraging 
transfer learning with ResNet50 and ensemble models for further 
accuracy improvements.