Neuron Classification Based On Their Morphological Features

The brain is an important organ that controls life activities.As the basic unit of the brain,neurons transmit various external information to the brain,which is the true core of information transmission.Therefore,the study of neurons is very necessary.The topic of geometric morphological features for the classification of neurons of different species is the basis of this research.However,from the current research status,this topic has not been well resolved.This paper mainly classifies the pyramidal neurons of human,chimpanzee and monkey by the morphological characteristics of neurons.Firstly,L-Measure software was used to extract the morphological features of pyramidal neurons of three species,and the statistical values of each feature were selected by searching related literature and related descriptions attached to the software.Secondly,quality analysis and feature analysis were performed on the data.Then,the two dimensionality reduction methods of PCA and LDA are used to reduce the extracted morphological features,and the new features after dimensionality reduction are analyzed respectively.Then,for each dimension reduction method,three types of classification algorithms(support vector machine,decision tree and random forest)are used to classify pyramidal neurons of three species respectively.Finally,the effects of different dimensionality reduction methods and different classification algorithms on the classification of pyramidal neurons are compared and analyzed.The results show that no matter which dimension reduction method is used to construct the classification model,the classification effect is ideal.After the dimension reduction based on principal component analysis,the support vector machine has the best classification effect,the random forest is slightly second,and the decision tree is the worst.In addition,the three classification algorithms are the best for the classification of human pyramidal neurons,followed by the monkey’s pyramidal neurons,the worst is the chimpanzee’s pyramidal neurons.After linearity discriminant analysis,the classification effect of support vector machine is slightly better than that of random forest and decision tree.However,support vector machine and decision tree have the best classification effect on chimpanzee pyramidal neurons,followed by the monkey’s pyramidal neurons,the worst is the human pyramidal neurons.After the dimension reduction based on principal component analysis,the classification effects of the three classification algorithms are better than those based on the linear discriminant analysis.And the classification model of the support vector machine is best in the classification model constructed by PCA and LDA dimension reduction methods.The study of these dimensionality reduction methods and classification algorithms can provide a basis for the classification of other neurons in different species based on spatial morphological features,and also deepen people’s understanding of the structure and function of neurons,and promote the development of neuroinformatics.