Paper Title
Deep Learning-Based Osteoporosis Detection From Knee X-Ray Images Using CNN Architectures

Abstract
Osteoporosis is a progressive skeletal disorder characterised by reduced bone mineral density and structural degradation, leading to increased fracture risk. Early detection of osteoporosis and its precursor stage, osteopenia, is essential for effective clinical intervention. However, standard diagnostic techniques such as Dual-energy X-ray Absorptiometry (DXA) are often limited by cost, accessibility, and availability, particularly in resource-constrained settings. In this study, a deep learning-based approach is proposed for the automated classification of knee X-ray images into three categories: normal, osteopenia, and osteoporosis. Two widely used Convolutional Neural Network (CNN) architectures, VGG16 and ResNet50, are employed to develop and evaluate the classification system. The models are trained using transfer learning and fine-tuned on a labelled dataset of knee radiographs. Standard preprocessing techniques and class-balanced training strategies are applied to enhance model performance. The proposed system demonstrates the feasibility of using deep learning for efficient and accessible osteoporosis screening with conventional X-ray imaging. The experimental results demonstrate strong classification capability, with VGG16 achieving the highest accuracy of 89.5% and superior overall performance. This work highlights the potential of artificial intelligence-assisted diagnostic tools in supporting early detection and improving clinical decision-making in osteoporosis management. Keywords - Convolutional Neural Networks, Deep Learning, Osteoporosis, Transfer Learning, Knee X-ray Imaging