Research On The Segmentation Algorithm Of Spinal Cord Scar

With the growing traffic accidents, injuries caused by sports and the accidents in mining industry, the incidence of spinal cord injury is on the increase. Studies have confirmed that the damaged nerve axons stop growing when contacting glial scar. If not treated effectively, spinal cord injury will be difficult to cure. However, if we can locate spinal cord scar quickly and accurately and determine its size, and cut the scar timely and effectively, the central nerve is likely to be cured. Therefore, the accurate and quick positioning and segmentation of spinal cord scar is critical to the recovery of spiral cord injury patient. As the rapid popularization and development of medical imaging technology, image processing technology and artificial intelligent technology, medical segmentation technology has become more intelligent and more automated, and is widely applied in the treatment of cancer, mutated and damaged tissues.At present, there are a large number of excellent medical image segmentation methods. This paper mainly conducts thorough research on MRI segmentation of spinal cord scar tissues, to put forward a fully-automated segmentation algorithm of spinal cord scar. This paper explains the significance of MRI spinal cord segmentation and the current research at home and abroad at first, and then makes analysis and summary on the key problems and major techniques at each link of MRI spinal cord segmentation technology. At last, the paper makes careful explanations on the segmentation method of spinal cord scar by MRI, and verifies the effectiveness of the proposed method through a series of experiments.The innovativeness of this paper can be summarized as follows:(1) In view of the characteristics of MRI and its noise problems, a self-adaptive non-edge information filtering smooth algorithm is put forward to separate the most common isolated pulse noise point and edge information point. And different pixels are processed by divergent filtering methods. First of all, isolated pulse noise points in MRI are screened, and then processed by smooth filtering technology. Afterwards, the edge information of image will be combined to filter non-isolated noise points by template weighted filtering method. This method can not only suppress current noise effectively in image, at the same time, it can keep effective edge information in the organization structure as much as possible, so as to make full use of the information.(2) Based on the layered relationships of MRI, deformable parts model is used to detect spinal cord injury for the first time, it can automatically locate and detect spine in section, thus greatly reducing the influence of noise generated by other tissues and organs when segmenting spinal cord scar, improving the efficiency of spinal cord scar segmentation algorithm, and in the meanwhile enhance the robustness of segmentation algorithm.(3) Based on coarse-to-fine strategy, a fully-automated spinal cord scar segmentation algorithm is put forward. This algorithm requires more interaction among personnel who conduct traditional medical image segmentation, which greatly increases the efficiency of segmentation, meanwhile, minimizes the personal emotion of operating personnel. This algorithm can narrow down the scope of detection gradually and combine the edge information image to ultimately detect the precise location of scar, so as to find out the location of seed point. Region-growing algorithm is also used to narrow down the scope of scar, and detect final segment scar accurately and effectively. Based on the comparative results of traditional medical image segmentation algorithm method and this proposed method, as well as false detecting rate and miss detection by stand segmentation method, this method is proved to be effective in the segmentation of spinal cord scar..