Accurate and clinically reliable detection of bone fractures in radiographic images remains a challenging task due to subtle fracture patterns, imaging variability, and the need for high sensitivity in real-world diagnostics. Existing deep-learning approaches often struggle to simultaneously achieve precise localization and robust classification, which limits their adoption in radiological workflows. To address these challenges, we propose HYEF-FD, a clinically oriented hybrid framework that couples high-sensitivity fracture localization with an ensemble-based classification strategy. The system first employs YOLOv11 to isolate fine-scale fracture regions, ensuring that subtle structural disruptions are accurately localized before classification. The extracted regions of interest are then analyzed by a soft-voting ensemble composed of ResNet50, DenseNet121, and EfficientNet-B3, leveraging their complementary strengths in fine-grained texture discrimination, medical-image feature extraction, and global representation learning. The code and related files can be accessed at https://github.com/HatemMajouri/HYEF-FD.git