Research On Spinal Localization And Segmentation Algorithm In MR Images

In recent years,spinal diseases have gradually become one of the major diseases affecting people's daily lives.The diagnosis of spinal diseases often requires the location of the markers,and manual labeling is time consuming and labor intensive.Research can automatically perform spine localization and segmentation algorithms on images of many different modalities,and the corresponding system will help doctors diagnose and treat diseases.For the positioning and segmentation of MR image segments,a strategy of gradually reducing the spatial extent of the image is employed.First find the most obvious layer of the vertebral block from the MR sagittal series of images,then determine the ROI region containing the vertebral block,and finally segment the vertebral block in the ROI region.When selecting the optimal layer,the vertebral area and the spinal cord area are synchronously presented,the gray area distribution of the spinal cord area is relatively uniform,the longitudinal strip is long,the gray level difference is large with the left and right sides,and the boundary with the spine area is clear.More easily than the area of ??the vertebral block,the most complete image of the spinal cord region is found as the optimal layer.Because the eight kinds of sequence images have different gray distribution characteristics in the spinal cord and spinal region,it is difficult to select the optimal layer by the same method.According to the left border length of the spinal cord and the gray distribution of the spinal cord region,the eight sequence images are divided into T1.,T2,Fat three categories.Different strategies are used for different types of images to select the optimal layer.In the ROI region of the spine,according to the characteristics of the spine region on the left side of the spinal cord region,the longitudinal strip shape and the width of the vertebral block are small,the right border of the spine is translated to the left as a reference line,and the longitudinal direction is searched on the left and right sides of the reference line.The boundary is screened and connected to obtain the left border of the spine region,and then the tail vertebra image is supplemented to obtain a complete spinal ROI region.The Fat-like spine image has the characteristics of large gray scale of the vertebralblock and dark gray of the intervertebral disc.The top hat transform and the bottom cap transform are used to increase the image contrast,and the OTSU algorithm is used to calculate the threshold value,and the image is threshold-divided to obtain the vertebral block.Segment the result.The spine region localization of MR images and the segmentation of Fat images in multiple modes are realized.According to the experimental data,the algorithm can effectively locate the spine regions of MR images and can segment the more accurate region of the vertebrae.