<p>Accurate segmentation of spine Magnetic Resonance Imaging (MRI) is essential for diagnosing conditions like low back pain and enhancing morphological studies. However, spine segmentation is challenging due to the complex anatomy, artifacts, and significant variability between slices. Traditional methods such as edge detection and machine learning techniques often require extensive manual tuning and struggle with the complex details of medical images. Recent advances in deep learning, particularly Convolutional Neural Networks (CNNs), have improved segmentation performance but face issues like high computational cost and overfitting especially with small medical image datasets. To overcome these challenges, we propose a simplified U-Net architecture tailored for lumbar spine MRI segmentation. This modified U-Net reduces model complexity while maintaining precise segmentation capabilities, which makes it suitable for applications with limited computational resources. Our approach adapts the U-Net structure to handle the specific characteristics of lumbar spine images and provides efficient and accurate segmentation with 94.7% of Dice Score. Experimental results demonstrate that our simplified U-Net offers high segmentation accuracy and can effectively support diagnostic processes in clinical settings.</p>