Lumbar degenerative spine disease is a prevalent age-related condition, particularly affecting older adults. It is primarily manifested through persistent lower back pain and progressive muscular stiffness, often leading to severe functional impairment and restricted mobility. A comprehensive pipeline was designed, beginning with the transformation of MRI-based DICOM images from the RSNA 2024 Lumbar Spine Degenerative Classification dataset into 3D voxel representations to enhance anatomical structure preservation. A Vision Transformer architecture was employed to model spatial dependencies and capture global contextual features more effectively than conventional convolutional approaches. The model was trained to classify degeneration severity across multiple categories relevant to clinical assessment. Experimental evaluation demonstrated strong predictive capability, with the model attaining a training accuracy of 98.7% (F1 score: 0.981) and maintaining robust generalization on the test set with an accuracy of 96.7% (F1 score: 0.970), indicating strong potential for real-world clinical application in supporting diagnostic decisions and improving workflow efficiency.