聚类分析在神经元形态特征分类中的应用

Application of cluster analysis in morphological characteristics of neurons

以49个神经元的三维坐标为基础数据,选取并计算出16个神经元的形态参数,然后对16个参数进行因子分析,选取出表征神经元大小、发散程度以及生长发育特征的3个特征因子,采用ward法对样本主因子进行系统聚类分析,从而对样本进行形态分类.结果表明:欧式距离为1.5左右时,研究样本可以分为运动神经元、双极中间神经元与锥体神经元、多级中间与三级中间神经元、感觉神经元以及普肯野神经元5类;神经元所处的发育程度对三级、多级与感觉神经元形态学分类有一定的干扰,因此,在形态学分类过程中,应使用发育成熟的神经元,避免造成干扰.

Base on the data for three dimensional（3D） coordinate of 49 neurons,16 morphological parameters were obtained by calculation.According to the factor analysis of the morphological parameters,three major characteristics,including dimension,degree of divergence and growths were obtained and chosen for classification.ward＇s method was applied in the system cluster analysis.The results showed that the neurons could be classified into 5 grades,including motoneuron,bipolar interneuron,cone neuron,multipolar interneuron and tripolar interneuron,when the Euclidean distance was about 1.5;additionally,the development degree of neurons affected the classification results of sensory neuron,tripolar interneuron and multipolar interneuron morphology,therefore,in the process of classification in morphology,it should be used mature neurons to avoid to cause interference.