Three Dimensional Reconstruction of Neuron Morphology from Confocal Microscopy Images

In recent years it has been more common to see 3D visualization of objects applied in many different areas. In neuroscience research, 3D visualization of neurons acquired at different depth views (i.e. image stacks) by means of confocal microscopy are of increase use. However in the best case, these visualizations only help to have a qualitative description of the neuron shape. Since it is well know that neuronal function is intimately related to its morphology. Having a precise characterization of neuronal structures such as axons and dendrites is critical to perform a quantitative analysis and thus it allows to design neuronal functional models based on neuron morphology. Currently there exists different commercial software to reconstruct neuronal arbors, however these processes are labor intensive since in most of the cases they are manually made. In this paper we propose a new software capable to reconstruct 3D neurons from confocal microscopy views in a more efficient way, with minimal user intervention. The propose algorithm is based on finding the tubular structures present in the stack of images using a modify version of the minimal graph cut algorithm. The model is generated from the segmented stack with a modified version of the Marching Cubes algorithm to generate de 3D isosurface. Herein we describe the principles of our 3D segmentation technique and the preliminary results.