基于数学形态学预处理的神经元放电检测  

Neural spike detection based on mathematical morphology preprocessing

作　　者：耿新玲[1] 乔志梅[1] 胡广书[1] 

机构地区：[1]清华大学生物医学工程系,北京100084

出　　处：《清华大学学报（自然科学版）》2009年第3期432-435,共4页Journal of Tsinghua University(Science and Technology)

摘　　要：神经元放电检测是对后续神经元放电波形的聚类分析、神经元放电串的统计分析等工作都非常关键的第一步。为了尽可能准确地把神经元放电从背景噪声中分离出来,该文首先应用数学形态学对原始记录数据进行预处理,然后再经过阈值处理把神经元放电检测出来。选择和放电波形在形状、持续时间和幅度上都相近的结构元素,数学形态学预处理步骤能有效地滤除原始记录中的背景噪声,突出放电信号。对神经元放电仿真数据集和实验记录的大鼠海马区神经元自发放电的检测结果显示:基于数学形态学预处理的神经元放电检测方法的准确率要高于常用的直接进行阈值处理的检测方法。The detection of neural spikes is an indispensable first step for the subsequent spike waveform clustering and spike train analyses. A spike detection method based on mathematical morphology preprocessing （MMP） of the raw extracellular recording was developed to accurately separate the spikes from background noise. By using a structure element with similar shape, duration, and amplitude to the spike waveforms, the MMP can eliminate most of the background noise from the raw data to make the spikes more evident, thus facilitating the following thresholding process. This method was tested by simulated neural spike data and records of spontaneous spikes fired by neurons in rat hippocampus. The results show a significant increase in the accurate detection rate compared with the usual amplitude thresholding method.

关 键 词：神经元放电检测 数学形态学 预处理 

分 类 号：Q424[生物学—神经生物学]