Segmentation of Vertebral Body Using Deep Learning Technique

In spine surgery, accurate insertion of pedicle screws plays an important role. Misplacement of the pedicle screw may damage the vertebral body and surrounding tissues, which can lead to nerve injury and paralysis. To avoid such complications during pedicle screw placement, the identification of the vertebral body structure is challenging due to its complex anatomy. Therefore, accurate identification of the vertebral body plays a crucial role in guiding the screw and preventing misplacement. Automatic segmentation of the vertebral body helps in developing patient-specific surgical templates along with pedicle screw guides, which assist the surgeon during the procedure. By placing the template onto the vertebral body, accurate positioning of the pedicle screw through the screw guide is ensured. The main goal of the current work is to develop a framework to automatically segment the vertebral body. Automatic segmentation is performed using a Convolutional Neural Network (CNN). CNN architectures are well suited for segmentation tasks as they can learn complex features and adapt to variations in patient anatomy and image quality. In this proposed work, automatic segmentation is performed using the U-Net architecture, which is well known for handling medical images that have image noise, anatomical variability, and inconsistent labeling. Accurate segmentation of the vertebral body can enhance the precision of pedicle screw placement and reduce complications during surgery.