This paper presents a new method for axis extraction from `tubular` structures in CT images. The axis is important in virtual colonoscopy systems because it can offer more efficient measuring features, interactive navigation, or other visualization possibilities. The proposed method is based on the `pseudo-tubular' structure of the Colon. To find a good estimation of the Colon axis, the method is organized in two steps: In the first step, the voxels are organized in sequential clusters along the structure using a region-growing strategy and a cilindricity criterium. Each cluster has a distinct point (which is a good candidate to be part of the axis, identified by the cilindricity criterium), and an internal partial ordering of the cluster´s voxels. This information structure allows us, in the second step, to propose three instances (one per principal coordinate axis) of an estimation problem to be solved with the Takagi-Sugeno-Kang Method (TSK). Based in the resulting solutions we were able to build a piece-wise linear estimation of the colon axis. The presented method has thrown good results. It is robust in face of the 'pseudo-tubular' shape of the colon. It has been tested on a set of Colon CT images from previously prepared patients. The results were illustrated with animations of virtual colonoscopy and a set of measurements and visualizations. Medical practitioners had a good perception of the method's potentiality in the field of virtual colonoscopy applications.