Microtubules are structural and motile elements. They are essential in numerous cell processes. Their function will be greatly improved by our understanding of their molecular structure, which requires extracting microtubules from the Cellular Electron Tomography Images. Manual segmentation and measurement of the microtubules from 3D dataset may require several man-hours of work, and the manual approach suffers from inter-observer and intra-observer variabilities.

Localizing microtubules axes is an important step for the successful segmentation. In this paper, a graph searching based technique for extracting central axes of microtubules from three dimensional cellular electron tomography images is presented. A tube enhancement filter is introduced in order to define a distance metric for graph searching. The results are promising, indicating that accurate mea-surements can be obtained in a semi-automatic fashion.