Algorithm Study For Disk Segmentation And Labeling Of MRI Spine Images

Spine MRI plays a very important role in diagnosing various spinal diseases like disk degeneration, scoliosis, osteoporosis and so on. Computer processing of spine MR images will make the diagnosis more accurate, consistent, time and labor saving. So, spine MRI processing is a significant work on clinical research.1. In this thesis, an unsupervised, automated and closed contours Marker Controlled Watershed segmentation algorithm was presented using in spine MRI segmentation. Firstly, morphological operations were used to preprocessing, to extract the background, enhance the object regions-disks, eradicated the disturbed regions; Secondly, we proposed a novel internal markers extracted algorithm based on the prior shape information which controlled the local minima regions numbers efficiently. The proposed algorithm can work on images with either disease. The experiments show the robustness of our algorithm.2. We proposed a semi-automated disk labeling algorithm based on the segmentation result. The method contents calculate the region centroids, refine searching region, formulate a criterion sets according to the statistic information. Find the real disks and labeling the medical name.All studies were supplied by the Neuroscience Institute, University Hospital, Cincinnati. The proposed disks segmentation algorithm achieved greater than 86% accuracy on 100 studies. The algorithm can get a very good result in the normal MRI spine, as well as in the scoliosis, disk deformable after spinal surgery.