Design And Implementation Of Shoulder Segmentation Algorithm In 3D CT Image

In the 3D CT human body images which include shoulder joints,the distances among the humerus,the scapula and the clavicle in shoulder joints are very close to each other,the traditional methods of bone segmentation make it apt to segment multiple bones into one;In addition,due to the differences of CT greyscale values between different bones and inside the identical bone,it is difficult to segment the shoulder joints and is apt to cause the multiple bones adhibited.Accurate separation of shoulder joints is of great significance for doctors to diagnose diseases.Grayscale threshold and Laplace filter threshold can influence regional growth between bones and inside the identical bone respectively.By setting higher initial grayscale threshold and Laplace filter threshold for each bone,and gradually decreasing the two thresholds to carry out the regional growth,the mutation of the number of pixel points in the grown region is detected and checked,by this way the two optimal thresholds for bone segmentation are dynamically determined,thus the segmentation threshold problem for multiple bones in shoulder joints is solved.This paper designs a new method to segment shoulder joints in 3D CT images.The proposed method divides the segmentation into three steps.In the first step,the lung is segmented,and then the seed points of bone contours of shoulder joints are found according to the position of the extracted lung.In the second step,according to the above-mentioned idea,the thresholds of grayscale and Laplace filter are determined dynamically,then the bone contour seed points gained in the first step are used to segment the contours of bones of shoulder joints by using region-based and edge-based segmentation methods.In the third step,for the contour of each bone segmented in the second step,the isotropic diffusion method and the closed operation are used to restore the corresponding bone.The shoulder joint segmentation tests were carried out by using 33 groups of 3D CT human body images with different number of layers and from different sources,the results show that the proposed method can accurately segment the bones of shoulder joints and the segmented bones can be restored into more acceptable images for doctors.