Segmentation of gallbladder from CT images for a surgical training system

A semi-automatic method was developed for the segmentation of 3D gallbladders (GB) from CT images, in order to construct a patient-specific model for a surgical training system. First a support vector machine (SVM) classifier was trained to extract GB region from one single 2D slice in the intermediate part of a GB by voxel classification. Then the extracted GB contour, after some morphological operations, was projected to the neighboring slices for automated re-sampling, learning and further voxel classification in these slices. This propagation procedure continued till all GB-containing slices were processed. The method was tested using 18 CT data sets and a set of quantitative measures were computed. The averaged volume overlap error of 15.56% and surface distance of 0.64 mm suggested that the method is efficient and promising.