Improving Medical X-Ray Imaging Diagnosis With Attention Mechanisms and Robust Transfer  Learning Techniques

Medical X-ray imaging plays a vital role in identifying abnormalities in bones and lungs. This project focuses on improving the accuracy of X-ray image classification using attention mechanisms and robust transfer learning techniques. Three datasets are utilized: Knee Fracture, Lung Cancer, and FracAtlas, covering different medical conditions. Pre-trained models are enhanced with attention modules such as Convolutional Block Attention Module (CBAM) and Squeeze-and-Excitation (SE) to extract relevant features and highlight important regions in images. Hybrid architectures combining deep learning models with attention mechanisms are developed for each dataset. The system is implemented using a Flask-based web application, enabling image upload and automated classification. The methodology includes preprocessing, model training, and evaluation to optimize feature learning. This approach demonstrates that integrating attention mechanisms with transfer learning improves the ability to detect subtle patterns in X-ray images. The study provides a framework for automated classification and contributes to the exploration of hybrid deep learning models in medical imaging.

Keywords: X-ray, attention mechanisms, transfer learning, CBAM, SE attention, ResNet50, DenseNet121, EfficientNet, ConvNeXt, VGG19