Multi-level Feature Enhancement and Dual AttentionbMechanisms for Improved Osteoporosis Diagnosis

In this paper, we proposed a novel deep learning model for osteoporosis detection that integrates dual attention mechanisms with multi-level feature fusion. Using EfficientNetV2 as the backbone and feature pyramid networks for multi-scale feature extraction, our model effectively captures deep semantic information and focuses on critical regions in knee X-ray images. Ablation studies and comparisons with state-of-the-art methods demonstrate that our approach improves classification metrics and reduces false positives, providing a more accurate and reliable tool for osteoporosis diagnosis.