Fuzzy Clustering Methods For Bone Segmentation In CT Images

Along with the development of medical imaging and the appearance of new technologies, a large numbers of image data were created by the new generation equipments of imaging and scanning technology; the demand for the softwares of 2D/3D image processing also fast grows. The increase of the information on diagnoses also proposed a higher demand for these processing tools with fast data analysing and accurate diagnosing ability. The algorithms which can extract the regions of interest or the anatomical structures from other backgrounds are one of the most important constituents of these processing tools .They're called image segmentation algorithms. The algorithms play an important role in numerous biomedical imaging applications such as the quantification of tissue volumes, localization of pathology, anatomical structure analysis, treatment planning and computer assistant diagnosis and so on. However, by reason of the complexity and the multiplicity of the medical images, there is no general segmentation algorithm to be used in all medical image processing presently, The segmentation algorithms used in the different clinical goals are still hot spots in the domain of medical image processing. Especially, the accurate segmentaion of bone which is an important tissue in human's body is playing a pivotal role of in many application domains such as treatment planning, assistant surgical operation and so on. However, the facts of the density largely varing and the weak edge effects make bone as one of the most difficult objects to be segmented.To get more accurate bone segmentation, two modified fuzzy clustering methods was proposed in this paper, which are formulated by modifying the objective function of the standard fuzzy c-means (FCM) algorithm .The two methods proposed base on the analysis of the character of the imaging of bone in CT images and the problems of the standard fuzzy clustering algorithm. One is the method that a penalty term is introduced into the object function of the standard fuzzy C-mean .With the addition of the adaptive penalty into the objective function, this method allows the adjustment of specific cluster scope according to the intensity variation of the cluster center during iterative process. The other is the method that two additional weighting factors are assigned respectively to each class and each voxel in the objective function. With the instruction of two weighting factors, the modified FCM has less sensitivity to the density and the size of cluster which standard FCM method suffers from. From...