In this chapter we presented a 3-dimensional multi-modal registration strategy for CT/MR images of the liver, with the aim of preserving physiologically plausible deformation at discontinuous borders. TV was used as the regularization term in place of traditional L2-regularization. We have demonstrated the robustness of the algorithm by experimenting with a 4-dimensional CT data set and clinical data. In the 4-dimensional CT data set, the TRE showed improved registration accuracy compared with some other state-of-the-art methods. The displacement field we obtained shows that TV can preserved discontinuation well at borders. In the clinical data, although we found little difference in the fusion image obtained with the two methods, the difference in the displacement field is obvious. In addition, better results were achieved for both HD and MHD with TV regularization than with L_2 regularization. All these results show that TV can preserved well real motion in the liver. The proposed method will be further improved in our subsequent work and be used in clinical applications.
展开▼
