Automatic and Reliable Extraction of Dendrite Backbone from Optical Microscopy Images

摘要

The morphology and structure of 3D dendritic backbones are the essential to understand the neuronal circuitry and behaviors in the neurodegen-erative diseases. As a big challenge, the research of extraction of dendritic backbones using image processing and analysis technology has attracted many computational scientists. This paper proposes a reliable and robust approach for automatically extract dendritic backbones in 3D optical microscopy images. Our systematic scheme is a gradient vector field based skeletonization approach. We first use self-snake based nonlinear diffusion, adaptive segmentation to smooth noise and segment the neuron object. Then we propose a hierarchical skeleton points detection algorithm (HSPD) using the measurement criteria of low divergence and high iso-surface principle curvature. We further create a minimum spanning tree to represent and establish effective connections among skeleton points and prune small and spurious branches. To improve the robustness and reliability, the dendrite backbones are refined by B-Spline kernel based data fitting. Experimental results on different datasets demonstrate that our approach has high reliability, good robustness and requires less user interaction.