Vertebral Block Segmentation Algorithm In MR Images

Spondylosis is a common disease.In the clinical diagnosis of spondylopathy,it is necessary to use multi-modal MR imaging technology to image the pathological part and segment the vertebral mass from the image.The study of automatic and high-precision vertebral block segmentation technology is helpful to the clinical diagnosis and follow-up treatment of spondylosis.The method of deep learning is used to segment the vertebrae.In order to improve the accuracy of segmentation,it is necessary to preprocess the input image and extract the spine region of interest,and reduce the impact of other organizations on segmentation.When extracting the spine ROI,according to the different characteristics of the spine region of different modes,the images of eight modes are automatically divided into three categories.According to the principle of the best integrity of the spinal region,the best layer image in the image sequence is found,that is,the most complete sagittal plane appears in the vertebral block.In the optimal layer image,different strategies are used for different types of images to obtain the right border of the spine region,and the image belongs to the body parts according to the right border of the spine.Image enhancement algorithm is used to improve image contrast,then the right border of spine is translated to the left as the reference line of the left border,and the candidate line of the left border is obtained near the reference line according to the edge detection results of Canny operator,so as to get the ROI of spine region.Using the tail vertebrae supplement algorithm,the tail region ROI of the image of the tail vertebrae is supplemented to obtain the complete spine ROI region.Finally,v-net neural network is used to segment the ROI region of spine to get the segmentation result.The algorithm of multi-modal and multi part spine automatic segmentation is realized and optimized after the algorithm is realized.The experimental results show that the algorithm can achieve the accurate segmentation of the vertebral block and has a short response time.