Axon Tracing And Postprocessing Of Neuron Morphological Reconstruction

Morphological reconstruction of neurons is critical for neuron classification,neuron circuits and studying the brain function.With the development of fluorescent labeling and imaging technology,people’s understanding of neurons has increased to the single cell level.Researchers have proposed various manual,semi-automatic and automatic neuron reconstruction algorithms and tools for neuron morphological reconstruction.However,due to the high diversity of neuron cells and the high complexity of images,it is difficult for the current automatic neuron reconstruction algorithms to reconstruct a large number of high-quality neurons because of insufficient generalization ability,especially in the case of neuron entanglement.In addition,although the current semi-automatic reconstruction can obtain high-quality neuron morphological reconstruction,it still has two shortcomings.First,the current semi-automatic reconstruction of neurons is a labor-intensive work,which is very time-consuming and labor-intensive.Second,there will inevitably be some quality problems due to the participation of people.Therefore,this paper has developed tools for neuron axon tracing and semi-automatic neuron reconstruction quality control and has proposed a pipeline of structure-based neuron morphology reconstruction automated pruning in the direction of reducing neuron reconstruction errors,improving neuron reconstruction quality,and accelerating neuron reconstruction speed.Details as follows:(1)An axon reconstruction algorithm based on the fast-marching algorithm was developed.This paper has designed an axon reconstruction algorithm based on fast marching algorithm,combined with local image information,and two-way verification,which effectively solves the reconstruction problem when neurons are entangled and improves the reconstruction efficiency of axon reconstruction.(2)A tool for quality control of artificial semi-automatic reconstructed neurons based on Vaa3 D was developed.The tool has formulated a standardized process,which can detect most of the hard quality indicators of neurons,greatly reducing the erroneous reconstruction of neurons,and effectively improving the quality of neuron reconstruction.(3)A structure-based neuron morphology reconstruction automated pruning pipeline(SNAP)was designed.For four different types of faulty reconstruction branches,SNAP has implemented single-neuron pruning and multi-neuron segmentation by incorporating specific statistical structural information into error branch detection rules integrated specific statistical structure information into the detection rules of faulty branches.Experiments have shown that SNAP is an effective post-processing tool for neuron reconstruction with high precision and high recall.