This research presents a comparative study of two deep learning models—AlexNet and ShuffleNet—for the automated detection and classification of spine abnormalities such as scoliosis and Spondylolisthesis. The objective is to evaluate each model's performance in terms of accuracy, precision, recall, and computational efficiency, which are critical factors in medical imaging applications. AlexNet, with its deeper architecture, achieved an accuracy of 91%, slightly outperforming ShuffleNet’s 82%, which is designed for more resource-constrained environments. Despite ShuffleNet's lower computational cost and faster inference time, AlexNet demonstrated superior accuracy and feature extraction capabilities, particularly in complex spine abnormality cases. This paper details the methodology, comparative analysis, and key findings, offering insights into the practical applications and potential improvements for deploying these models in clinical settings.