3D dendrite reconstruction and spine identification.

In neuron-biology, 3D neuron dendrite reconstruction followed by spine identification is indispensable for the study of neuronal functions and biophysical properties. In this paper, we propose an automatic dendrite reconstruction method to with a surface representation of the neuron on the basis of a novel level set approach. Our novel level set approach can effectively tackle the problem of segmentation under blurring and intensity in-homogeneity. Then spines are detected based on dendrite medial axis and a label-based thinning strategy is proposed to accurately extract the dendrite skeleton for spine identification. Experimental results reveal that our method works well.