Abstract: With the aging population and the increasing prevalence of fracture-related diseases,timely and accurate diagnosis of spinal fractures has become increasingly important.Although traditional imaging modalities,including X-ray,computed tomography(CT),and magnetic resonance imaging(MRI),remain the gold standards for diagnosis,they are challenged by complex spinal anatomy and the similarity between adjacent vertebrae,which frequently lead to a certain proportion of missed diagnoses.In recent years,deep learning has demonstrated substantial potential in medical imaging,particularly in the diagnosis of spinal fractures,where it has improved detection rates and diagnostic accuracy.This review summarizes recent advances in deep learning-based imaging diagnosis of spinal fractures,covering key applications such as automated detection,image segmentation,fracture classification,and preoperative planning.In addition,current challenges are discussed,including issues related to data quality,limited model interpretability,barriers to clinical translation,and technical limitations.To facilitate the clinical implementation of artificial intelligence technologies,future research directions are proposed,including innovative technologies such as few-shot learning,federated learning,and multimodal data fusion.Ultimately,the continued advancement of deep learning techniques is expected to enhance personalized and precise diagnostic capabilities as well as prognostic management for patients with spinal fractures.

Key words: deep learning, spinal fracture, convolutional neural network, artificial intelligence