文摘
Image segmentation of the region of interest is important in computer-integrated medical intervention. The segmented results are used to reconstruct 3D model of the organs of interest, and align the pre-operative image data with the physical space of a patient in image-guided medical interventions. Pre-operational surgery plans are based on the accuracy of the geometry of 3D reconstructed models. Registration of the model gained from CT scans must accurately match the three dimensional bones from which the surgeon will work. Thus, accurate segmentation is desired. However, we have found it is very difficult to automatically segment bone precisely, especially at joints, due to injuries, bone loss, bone's inhomogeneous structure, and the limitation and resolution of CT images. Under this circumstance, indication of regions of pathological shape change becomes extremely important to enhance the automatic segmentation and correct registration. We employ Fourier descriptors as bone shape signature. It is followed by a statistical analysis to detect the region of abnormal bone shape variation.