The goal of this study is to develop a novel Deep learning (DL) AI model, utilizing convolutive neural networks (CNN) and transformer architectures, to automate the detection and classification of cervical spine compression. Two musculoskeletal radiologists labeled a total of 795 cases of cervical spine MRIs, followed by classification using the Kang system by a separate group of six clinicians. The model was then trained on detecting regions of interest and grading capability using the Kang system. The model achieved an AUC (area under the receiver operating characteristic curve) above 0.930 across all evaluated categories, including diagnosis, grading, stenosis, and significance. When compared to the performance of clinicians, the model was automatically able to assess cervical spine pathologies similarly. This study’s primary drawback is that mild-to-moderate spinal stenosis cases were used, leading to issues in generalizability further down the line for extreme cases.