Prognosticating Pneumonia: A Comparative Analysis of Prediction Models Enhanced by and Devoid of Attention Mechanism

Pneumonia is a serious condition that requires prompt diagnosis and intervention. Chest X-ray being the primary tool for identifying pneumonia, allows doctors to detect the early onset of the disease to ensure effective treatment and prevent complications. Chest X-ray is chosen in our study owing to low cost and convenience as opposed to other imaging techniques. Conventional diagnostic approaches, while reliable, can be labor-intensive and require significant human resources. To overcome this limitation, implementing deep learning algorithms for automated pneumonia detection can significantly accelerate the diagnosis process. The primary contribution of this research is to employ attention mechanism in a custom and an adapted convolutional neural network model, to evaluate how attention mechanism affects the classification accuracy and other metrics. The data sets of chest X-ray images are injected into the deep learning models to accurately classify the images as either pneumonia or normal cases. Augmentation techniques, such as rotation, and contrast adjustment are utilized to handle the class imbalance in the dataset. The results demonstrate a notable improvement in pneumonia detection accuracy when the attention mechanism is applied, highlighting its potential to enhance deep learning models in medical imaging analysis. This study underscores the efficacy of attention-enhanced CNN models for reliable, automated pneumonia detection, contributing to the broader effort of advancing diagnostic tools through machine learning.