Tools
Tools
Koli, Manav Siddharam (2024) Enhancing Diagnostic Accuracy: AI-Enabled Chest X-ray Analysis for Improved Pneumonia Detection. Masters thesis, Dublin, National College of Ireland.
Preview |
PDF (Master of Science)
Download (1MB) | Preview |
Preview |
PDF (Configuration Manual)
Download (1MB) | Preview |
Abstract
Pneumonia remains a leading cause of mortality, particularly among young children and the elderly, underscoring the urgent need for accurate and timely diagnosis. This study offers a novel method for using artificial intelligence (AI) to improve the precision of pneumonia diagnosis using sophisticated chest X-ray processing. We used Convolutional Neural Networks (CNNs) to create a reliable AI system, making use of cutting-edge models such as ResNet152V2 and VGG16. In order to guarantee a balanced dataset, our approach comprised rigorous data preprocessing such as normalisation, scaling, and data augmentation. Although, using Hyperband for hyperparameter adjustment improves our model training outcomes. Grad-CAM (Gradient-weighted Class Activation Mapping) is a novel approach in our work that not only improves model transparency but also offers important visual insights into the particular regions of chest X-rays that show pneumonia. This feature greatly improves diagnosis reliability by enabling medical practitioners to precisely pinpoint impacted regions. Our models demonstrated exceptional accuracy, achieving 94.23% with VGG16 and 92.26% with ResNet152V2, highlighting the effectiveness of our approach. This research not only advances the field of AI in medical diagnostics but also offers a transparent, interpretable, and highly accurate tool for pneumonia detection, with the potential to transform patient outcomes, especially in resource-limited settings.
Actions (login required)
 |
View Item |