3-D centerline extraction of axons in microscopic stacks for the study of motor neuron behavior in developing muscles

This paper presents a new algorithm for extracting the centerlines of the axons from 3-D data stacks collected from a laser scanning confocal microscope. Recovery of neuronal structure from such datasets is critical for quantitatively addressing a range of basic biological questions such as the manner in which the branching pattern of motor neurons change during synapse elimination. The presence of artifacts in the crosssectional images such as blurred boundaries and non-uniform intensities, makes the process of centerline extraction rather challenging. Although many methods exist in practice today, they are either error-prone or involve manual interaction to a large extent, when applied to this particular problem. We propose a robust probabilistic region growing algorithm to extract the centers from the datasets with minimal user interaction.