Detection and classification of spinal tumors remain a challenge due to the lack of available annotated data. This motivated us to create an augmented database for researchers to use. This study introduces a novel data augmentation technique in which we insert tumors along the cerebrospinal fluid in healthy lumbar spine MRI scans. This method addresses the scarcity of labeled data by generating anatomically realistic tumor variations, significantly enhancing training diversity and model robustness. We have systematically generated the augmented data with 100% accuracy in augmentation. Using this dataset, a fully automated pipeline achieves 99% accuracy in classifying spinal tumor types from T2-weighted MRI images. The system has a 6-layer Convolutional Neural Network (CNN) strategy and incorporates anatomical priors for context-aware predictions. This augmentation-driven approach offers a clinically relevant solution for spinal tumor diagnosis.