Image segmentation using MRI vertebral cross-sections

Computer assisted spinal surgery requires the development of a 3-D image of a patient's spine. A method is being developed to construct such a 3-D spinal image from axial MRI cross-sections, using a deformable template. This paper outlines techniques used to register the model with the patient data. First an algorithm based on symmetry of the anatomy is developed to determine the position and orientation of the vertebrae in each cross-section. Second, an active contour is used for segmentation of the spinal cord. Last, a novel edge detection scheme is developed to identify the low contrast edges of the vertebral body in each cross-section.