A semi-automatic method for neuron centerline extraction in confocal microscopic image stack

Abstract

An algorithm for extracting the centerlines of neurons from 3-D image stack collected from a laser scanning confocal microscope is presented. Recovery of neuronal structure from image stack is critical for quantitative analysis of neuron-morphology. Many methods have been proposed to extract the centerline from the tubular structure in medical images, such as vessels. But the same methods do not work well in processing of neurons. One of the reasons is that the physical limitations of the spatial resolution of the image stack collect by confocal microscopy in the z-direction is much worse than the resolution in x and y directions. In our studied cases the mean voxel size of the image stack is 0.17 x 0.17 x 1.0 mu m(3) i.e., the resolution in z-direction cannot reflect the fact that the neuron has a tubular structure. In this paper, we propose an almost automatic neuron-tracing algorithm for a set of confocal microscopic images of neuron. The method is designed based on finding a minimal path in the volume to extract the centerlines of a neuron.