We report a method for semi-automated segmentation of extended features such as filamentous structures in electron tomograms. We present an application of this method for the automatic propagation of segmented neuro-filaments present along the axis of an axonal section of fixed, plastic-embedded mouse brain tissue imaged by three-dimensional electron microscopy. The algorithm uses the manually segmented features in a given slice to progressively extend the segmentation to successive slices based on knowledge of the centroids, shapes, and intensities of filaments in the previous slice. An iterative procedure allows for refinement of the most probable locations of the filaments in each slice. Evaluation of the performance of the propagation algorithm by visual inspection suggests that this approach can be used successfully for the rapid quantitative and statistical analysis of filamentous structures in cellular tomograms.