Objective <p>This study explores the application of deep learning in the classification of rib fracture injury mechanisms based on medical images. It aims to address the challenges of identifying rib fracture injury mechanisms and the complexity of network architectures across different deep learning models, thereby determining the most effective model for this task. Five deep learning models were evaluated for their performance in classifying rib fracture injury mechanisms. CT medical datasets containing rib fractures of various types and morphologies were adopted, and image preprocessing techniques such as normalization and scale adjustment were used to improve model performance. Each convolutional neural network (CNN) model was carefully designed, trained, and tested, with a focus on accuracy, sensitivity, specificity, and computational efficiency.</p> Result <p>Comparative analysis of the models revealed that each architecture had its own advantages and disadvantages in the classification of rib fracture injury mechanisms, with significant differences in overall performance. The key finding was that all models accurately detected rib fractures and distinguished between different injury mechanisms.</p> Conclusion <p>This study focuses on the clinical value of different CNN models for automated detection of rib fracture mechanisms, providing practical evidence for the intelligent upgrading of medical image analysis. The results demonstrate that the ResNet18 model achieves the best performance on this task, with an overall accuracy of 95.44%, strongly validating the potential of deep learning to improve diagnostic efficiency. With strong practical implementation capabilities, this model can be integrated into the CT pre-screening workflows of forensic institutions and public security systems in the future and also deployed on lightweight mobile or edge diagnostic devices, facilitating rapid and accurate assessment of rib fracture injury mechanisms in resource-constrained settings.</p>