Since the advent of computed tomography, methods for automatically detecting anatomical objects in CT and MR images have been active areas of research in the medical imaging community. Most of the published work has dealt with the application of such techniques to cranial images, and very little has dealt with the area of spinal images. The reasons for this are manifold. The anatomical scale of spinal elements is small compared with the field of view of the images. As a result, the spatial contrast of the features of interest is very poor and the features become very noisy and ambiguous. In addition, the topological characteristics of various spinal elements, such as the vertebral bodies and facets, are more complex than many of those found in the cranium. Finally, relative motion among the vertebral bodies can invalidate certain techniques based on rigid-body assumptions that are applicable to the cranium.
This chapter is intended to provide a review of existing image understanding methods and provide a motivation for the design of the feature-recognition techniques used in my research. I will start by reviewing some of the common approaches used in image understanding and explain the strengths and limitations of each approach.