Active Contour Model Based On The Open Curve Of 3 D Nerve Target Segmentation And Central Axis Extraction Algorithm

Reconstructed From the3D neuron image is a hotspot of modern neurobiology. and brain’s function is closely related to neuronal morphology and structure. We can understand how the brain work, and prevent some neurological diseases by study neuron. While the key to the neuron reconstruction process is to correctly extract the central axis of the neural target. it’s very complicated for us to using manual identification deal with3D fluorescence confocal microscopy image sample. and extraction accuracy is difficult to meet the requirements. so we need to study the new methods with computer that help automated extraction and detection of neuron. this paper focuses on3D neuron segmentation and centerline extraction method based on open-curve snake and proposed a systemic solution. which includes a series of pre-process and post-process methods.The main work in this article shows as follows:(1) Introduced the theory and construction methods of image enhancement which is based on the Frangi’s vesselness measure model, and proposed an improved model which is simpler。Using this simplified model do enhancing and denoising with the original image,finally obtained an vesselness image.(2) Combining with vesselness measure to construct the energy term,a pre-segmentation is applied on the original image using graph cuts method. Using this method successfully divided the original image into object and background regions.(3) two types of seeds detect methods was studied on neuron image,one of them is a classic method which was based on ridge criterion, the second method was based on linear scan done with the pre-segmented image, the experiment of this algorithm showed that it is less time-consuming and could obtain more valid seed points.(4) Researched the process of centerline extraction based on open-curve snake tracing model, the open-curve was started with the valid seed points detected before. During the evolution,combined with the GVF which is employed as a deforming force and the stretching force.so that the open-snake can be stretched bi-directionally along the direction of axon centered at the centerline. This article’s contribution was we have given two types of stretching force,one is based on the main eigenvector of the Hess ion matrix and the other is based on vesselness measure. Experiments showed a comparison of extraction results with these two types of extended force.(5) Proposed an automatic3D confocal image of neuron centerline extraction program based on the above process, and provided a modular process design.Experiments show its effectiveness.