Spine fractures pose a critical health risk, demanding precise segmentation for effective diagnosis and treatment. This research introduces MSFF, a Multi-Scale Feature Fusion deep learning model for automated spine fracture segmentation in MRI images. MSFF integrates five key modules—FFM, SEM, ASPP, RCBAM, RBRAB, and LPRAB—to enhance multi-scale feature fusion, spatial and channel-wise feature extraction, border refinement, and positional attention. A decoder network then predicts fractures. Experimental results show that MSFF outperforms existing methods in accuracy and segmentation quality.