||I will present a method for segmentation of the liver parenchyma from CT images. The goal of the segmentation is to measure liver volume and metastasis volume for a patient at two different times. The segmentation has been performed on 52 datasets (26 patients) by letting two users place seed regions inside the liver using a 3D interaction environment with haptic feedback. From the initial regions a surface is propagated by using a Fast marching algorithm with a cost function based on gradient magnitude and intensity. The segmentation results contain errors due to leaking, and I will present some ideas on how the result can be refined by using deformable surfaces.