<p>Accurate fracture diagnosis from radiographic images remains challenging due to its heavy reliance on manual interpretation, which is susceptible to variability in clinical experience, imaging quality, and workload fluctuations. To enhance detection reliability and efficiency, this study introduces ABCP-YOLO11, an improved fracture detection framework built upon the YOLO11 architecture. The model integrates multiple complementary attention mechanisms–Channel Attention, Adaptive Regional Attention (AAttn), and Position-Sensitive Attention (PSA)–to strengthen feature representation across channel, spatial, and positional dimensions. Combined with diverse augmentation strategies, these modules enhance the recognition of subtle cortical disruptions and fractures embedded in complex anatomical structures. Experiments on multi-source X-ray datasets demonstrate significant improvements over the baseline YOLO11, including a 4.6-point increase in mAP@0.5 and a 5.6-point rise in recall, while maintaining real-time inference at approximately 50 FPS. Five-fold cross-validation confirms the robustness and statistical significance of these improvements over the baseline. Overall, ABCP-YOLO11 provides an effective and clinically meaningful solution for automated fracture detection.</p>