Segmentation of liver and liver tumor for the Liver-Workbench

摘要: Robust and efficient segmentation tools are important for the quantification of 3D liver tumor volumes which can greatly help clinicians in clinical decision-making treatment planning. A two-module image analysis procedure which integrates two novel semi-automatic algorithms has been developed to segment tumors from multi-detector computed tomography (MDCT) images. The first module is volume using a flippingfree mesh deformation model. In each iteration, before mesh deformation, algorithm detects avoids possible flippings which will cause self-intersection then undesired results. After flipping avoidance, Laplacian performed with various constraints geometry shape smoothness. the second module, segmented used as ROI by support vector machines (SVMs)-based voxel classification propagational learning. First SVM classifier was trained extract tumor region from one single 2D slice intermediate part classification. Then extracted tumor contour, after some morphological operations, projected its neighboring slices automated sampling, learning further slices. This propagation procedure continued till all tumorcontaining slices were processed. performance whole tested 20 MDCT data sets the results promising: Nineteen successfully out, mean relative absolute volume difference (RAVD), overlap error (VOE) average symmetric surface distance (ASSD) reference segmentation 7.1%, 12.3% 2.5 mm, respectively. For live segmentation, median RAVD, VOE and ASSD 7.3%, 18.4%, 1.7