In this paper, we present a method for automated extraction of the liver region from contrasted 3D abdominal CT images. Since the medical doctor diagnoses the liver using multi contrasted phase CT images, it is necessary for computer aided diagnosis system which extracts the liver region from each contrasted phase image. We propose an automated extraction method of liver regions from both the early and late arterial phase CT images. In the case of the late arterial phase, we estimate the threshold value for the liver region extraction by analyzing the CT value histogram. Then, false positive regions are removed by employing a figure decomposition and composition method based on an Euclidean distance feature. Finally, we modify the contour of the liver region. In the case of the early arterial phase, we map the liver region extracted from the late arterial phase onto the early arterial phase CT image and modify its contour. We applied the proposed method to contrasted abdominal CT images. The experimental results suggested the proposed method could extract the liver region from both phases satisfactorily.

View full abstract