Sparse Principal Component Analysis Of Neuronal Morphological Feature Extraction And Its Characteristics

Neurons as the basic unit of the brain,its shape is complex and diverse.This feature of neuronal morphology plays a vital role in the normal functioning of the brain,so it is important to understand the differences between the morphological characteristics of different types of neurons and to understand why the brain has such complex functions and sophisticated structures.The geometric characteristics of neurons can represent the type and structure of neurons to a large extent.Therefore,the study of neuronal type and structure requires automatic extraction of neuron geometry and effective statistical analysis of extracted features.At present,there are dozens of geometric features describing neurons,and there are related tools that automatically extract these geometric features.However,these tools are scattered,closed,and difficult to integrate with other neuronal morphological reconstruction platforms.In this case,this paper designs and develops a software tool that automatically calculates the geometric characteristics of neuron morphology.The software tool is open source and can be combined with neuronal morphological reconstruction software tools.On the basis of the automatic extraction of the geometrical features of the neurons,the software tools developed also provide statistical analysis of the geometric features.Considering that the number of geometric features of neurons is too large,and many geometric features have strong correlation.When analyzing different types of neuron morphological differences or when recognizing and sorting neurons,what kind of geometric features are selected and how much geometric features are selected is a more difficult problem.In order to solve this problem,it is necessary to adopt the appropriate statistical method to obtain the key features of the description form from many morphological features.Based on this demand,the introduction of sparse principalcomponent analysis,to integrate the current number of morphological features of the information,the formation of several key several morphological features.The five neurons in NeuroMorpho.Org database were analyzed by using self-developed tools and the sparse principal component method.The results are summarized as follows.Each type of neuron has a similar principal component,such as the branching of the neurons and the junction of the neurons,the overall contours of the neurons,the connections of the neuronal branches and the trunk,and the neuronal elongation Et al.,But each type of neuron is composed of different types of principal component combinations,have a unique principal component,which is directly related to the morphological characteristics of different types of neurons.