GNMR:3D Neuron Morphology Retrieval Based on Graph Neural Network
Neuron morphology and structure is an important task to analyze neuron activity and development function.How to effectively identify neurons of different shapes is a challenge.This paper proposes a new method of 3D neuron morphology retrieval based on graph convolutional neural network(GNMR for short).First,the child-parent node scheme is used to preprocess the three-dimensional neuron.According to the spatial geometric structure of the three-dimensional shape,a three-dimensional neuron is mapped to the three planes of X-Y,X-Z and Y-Z.Secondly,a GNMR is designed to retrieve Neuron shape,in order to avoid the problem of gradient explosion and gradient disappearance,three layers of ReLU function are added to the connection layer.Finally,the method is simulated on the NEU-1500 data set.The experimental results show that the method can effectively identify the shape of three-dimensional neurons,and has high retrieval accuracy,precision and recall.
neuron geometrygraph neural networkneuron recognitionReLU function
国家科技期刊平台
NSTL
万方数据
