Lumbar spine degeneration is a leading cause of chronic lower back pain, affecting millions globally. This study presents a deep learning-based approach using Convolutional Neural Networks (CNNs) to automatically classify the severity of lumbar spine degenerative conditions from MRI images. Leveraging the RSNA (Radiological Society of North America) Kaggle dataset of 147,000 images, we implemented and evaluated multiple CNN architectures, including ResNet50, DenseNet121, and VGG19, combined with structured metadata. Our DenseNet-based model achieved over 75% validation accuracy, highlighting the potential of automated tools to enhance diagnostic precision in radiology. The approach shows promise for supporting radiologists in clinical decision-making and reducing diagnostic subjectivity.