Abstract: Fusion of CT urography images is necessary in kidney preoperative planning system. A kidney-centered registration is the first step in the fusion. In this paper, several steps were taken to achieve this goal. Semi-automatic segmentation based on DDCM model was first performed to segment the kidney contours from CT slices. A binary kidney volume was obtained by contours reconstruction. The binary volume was then intersected with the original gray volume after 3D dilated. Finally, rigid registration between the kidney gray volumes was accomplished. The effect of initial parameters during registration was analyzed. Characteristics of geometric moments were applied to initialize parameters in order to avoid local extrema. Experimental results show that the proposed method can realize a kidney-centered registration between CT acquisitions, and the precision and speed can meet practical requirements.

Key words: nephron-sparing surgery, preoperative planning, 3D registration, mutual information