Algorithm Study For Disk Labeling And Vertebral Segmentation Of MRI Spine Images

Medical image processing is a kind of technique that makes use of image vision to assist clinical diagnoses. Therefore, it has become a research hotspot nowadays. The study of the segmentation of MRI Spine Image is of crucial importance for computer aided medical image identifying and clinical studies of neurological pathology. Computer characterization of a vertebra or disk is of limited clinical value if that structure cannot be accurately segmented and identified. However, Manual operations of tracking these structures are so tedious that automated method of spine tracking and segmentation are in high demand.In this paper, both disk labeling and vertebral segmentation algorithm are proposed. After pre-processing of the image slices, model-based method is used to realize the rough segmentation of the spine region, and then the spinal cord is located using Hough Transform. Along the curve of spinal cord and under some rules of labeling, we are successfully labeling every disk in the whole spine. And the accurate locating the coordinates of the disks provides a good initialized position for vertebral segmentation. Active Shape Model (ASM) performs excellent in the vertebral segmentation. The efficiency of the proposed algorithm is demonstrated by experiments using real MR images provided by College of Medicine, University of Cincinnati. And we also propose further improvement in this study.In the end, we implement the medical image processing in a system based on FPGA. Meanwhile we transfer the Disk Labeling algorithm to an IP core which makes this technology be easily used in lots of fields.