Shape-constrained repulsive snake method to segment and track neurons in 3D microscopy images

Abstract

To study the structure and branch pattern of neurons, it is important to segment and track neurons at first. We develop a snake model based on repulsive force to segment neurons in 3D microscopy image stacks. To overcome the difficulty that the boundary between two adjacent neurons is weak, we introduce a shape constraint on the snake deformation and use repulsive force to keep snakes of adjacent objects from merging into one. After obtaining the contours on the first image slice, we project them to the next slice as initialization for snake deformation and repeat the process for all the slices in a 3D image stacks. Individual neuron is tracked by connecting the corresponding snake through all slices. Results obtained from processing real data show that the method can successfully segment two or more neurons that are close to each other in 3D. © 2006 IEEE.