Automatic Lumbar Vertebrae Segmentation in Fluoroscopic Images via a New Parallel Hough Transform

Low back pain is a very common problem and the consequent cost is enormous. In spite of the prevalence, however, it is still difficult to obtain accurate diagnostics. Mechanical disorder is now often regarded as the main cause. Presently, a number of studies are directed at measurement of spine kinematics. Since the lumbar spine is a deep-rooted structure, this has been difficult until digital video fluoroscopic (DVF) imaging was developed. In this paper, DVF is used to acquire the motion sequence of the lumbar spine, to solve the problems with manual location of the vertebrae within the sequence; a new real-time and parallel Hough Transform hardware algorithm has been developed to automatically locate vertebrae in DVF images. Experimental Results indicate that Lumbar Vertebrae Segmentation can be achieved in frame rate, using low amount of hardware. Key-Words: Image Segmentation, Digital Video Fluoroscopy, Real-time processing, Hardware design.