This paper presents improvement of a segmentation procedure of kidneys using deformable model and its performance evaluation. In our early work, we have developed a kidney segmentation method based on a 3D deformable model, which uses statistical information about the organ's shape. Using the proposed method, we have extracted left kidneys from 3D abdominal CT images whose slice interval is 10mm and evaluated the performance. However, automated positioning of initial models and application to right kidneys remained to be done. In this paper, we propose automated locating process of initial models in 3D images. Moreover, the experimental results of segmentation of both left and right kidneys are also given with the discussion on effectiveness of the proposed procedure.

View full abstract