Reconstructing the Surface Mesh Representation for Single Neuron.

In this paper, we present a pipeline to reconstruct the membrane surface of single neuron. Based on the abstract skeleton described by points with diameter information, a surface mesh representation is generated to approximate the neuronal membrane. The neuron has multi-branches (called neurites) connected together. Using a pushing-forward way, the algorithm computes a series of non-parallel contour lines along the extension direction of each neurite. These contours are self-adaptive to the neurite's cross-sectional shape size and then be connected sequentially to form the surface. The soma is a unique part for the nerve cell but is usually detached to the neurites when reconstructed previously. The algorithm creates a suitable point set and obtains its surface mesh by triangulation, which can be combined with the surface of different neurite branches exactly to get the whole mesh model. Compared with the measurements, experiments show that our method is conducive to reconstruct high quality and density surface for single neuron.