Research And Implementation Of Neuronal Structure Reconstruction Algorithm Based On Local Radius Estimation

Confocal light microscope is used to observe and collect the images of neuronal cells in the brain tissue.Afterwards,we can design the model reconstructed the three-dimensional morphological structure of neuronal cells.And the visual interaction analysis of the structural reconstruction results can help researchers understand and reveal the brain's function and behavior mechanism.This research method is one of the most important methods of neuroscience research.It can accurately record the morphological structure of neurons and reduce the storage capacity of neuronal structure data.The reconstruction of neuronal morphological structure is divided into the following steps:(1)Through the analysis of the three-dimensional neuronal image structure,the skeleton of the neuron is extracted;(2)The neuronal local radius of the node on the skeleton is calculated.It can be seen that the local radius of the neuron can accurately describe the local branch structure of the neuron and play a crucial role in the structure reconstruction of the neuron.The main research content of this paper is the estimation of neuron local radius in 3D neuronal images.Then based on the study of the local radius estimation of neuron,the neuron structure reconstruction algorithm is studied and implemented as software.The work is of a great implications for the scientific research of neuronal structures.The main work of this article is divided into the following points:First of all,this paper propose a structure analysis algorithm based on adaptive Rayburst model,which can be used to estimate the local radius of three-dimensional neuron images:first select the node on the neuronal skeleton in the three-dimensional image and analyze the pixels around the node and the neuron structure information.Afterwards,the integral image is calculated to obtain the adaptive threshold at the node.Then,a Rayburst model is established at this node,and the surface pixel intensity threshold in the Rayburst model is adjusted according to the calculated adaptive threshold,and the threshold is used to distinguish the foreground from the background region to terminate the growth of the ray.By statistically analyzing the length of each ray in the Rayburst model,an estimate of the local radius of the neuron can be obtained.Compared with the original Rayburst model,the adaptive Rayburst model proposed in this paper can effectively avoid the influence of the noise region and the uneven pixel intensity region on the local radius estimation of neuron in the three-dimensional neuron image.Afterwards,neuron local radius estimation algorithm is combined with neuronal terminal point detection algorithm and all path pruning algorithm to reconstructe the neurons in 3D neuron images.At the same time,the neuron structure reconstruction algorithm introduced in this paper is implemented as a software,which can complete the fully automatic neuron structure reconstruction.Finally,the local neuron radius estimation algorithm proposed in this paper is applied to the three-dimensional image of the drosophila olfactory neuron cell in the classical DIAMEM challenge and BigNeuron engineering database.Experimental and data analysis verify that the local radius estimation algorithm proposed has extremely high accuracy and robustness.At the same time,the visual analysis also shows that the software implemented in this paper can be used in the actual neuron structure reconstruction of three-dimensional neuron image,and can get good results.