Abstract: Cervical spine fractures are a critical medical emergency that can lead to severe com-plications, including permanent disability or death if not diagnosed promptly. A cervi-cal spine fracture may be detected by using computed tomography (CT). This study pre-sents a deep learning-based approach for the classification of cervical spine fractures us-ing a dataset containing computed tomography (CT) images of fractured and normal cervical spines. The proposed methodology incorporates transfer learning models, in-cluding DenseNet121, VGG16, and MobileNet, to achieve high accuracy in distinguish-ing between normal and fractured cervical spines. The study evaluates model accuracy, precision, recall, and F1-score to determine the most effective architecture. Experimental results indicate that the VGG16 model optimized with the Nadam optimizer achieves the highest classification accuracy of 98.37%, outperforming other models or the same model with another optimizer. The findings highlight the potential of deep learning in assisting radiologists with faster and more reliable cervical spine fracture detection, ul-timately improving patient care and reducing diagnostic delays.