采用基于密度加权和偏好信息的K均值聚类的胸阻抗信号自动检测算法

Automatic Detection Algorithm for Transthoracic Impedance Signal Using K-means Clustering Based on Density Weighting and Preference Information

为了自动识别胸阻抗(Trans Thoracic Impedance,TTI)信号中的按压和通气波形,完成相关重要参数的计算,从而实现对心肺复苏质量的监测评估,该文提出一种基于密度加权与偏好信息的胸阻抗信号自动检测算法。该方法针对实验采集的猪的电诱导心脏骤停模型TTI信号,通过预处理和多分辨率窗口搜索法完成潜在按压和通气波形的标记;接着,提取每个标记波形的宽度、幅值以及相邻波形特征差作为特征,并按标记波形宽度对信号进行分段;之后,再对信号进行小波分解,提取其小波系数每段的能量与原始波形幅值之比作为特征;最后采用基于密度加权与偏好信息的K均值聚类分析法对标记的波形进行分类识别。实验结果表明,该算法对TTI信号中按压波形和波形分析识别的正确率和敏感度均较高,鲁棒性好,且运行时间(0.43 s±0.07 s)满足实时性要求。

In order to recognize automatically the compression and ventilation waveforms of the Trans Thoracic Impedance（TTI） signal, and obtain the important parameters, for evaluating the Cardio Pulmonary Resuscitation（CPR） quality, this paper proposes an automatic detection algorithm for TTI signal based on density weighting and preference information. The TTI signals that come from the pig model based on electrically induced cardiac arrest are preprocessed, and the potential compression and ventilation waveforms are marked by using the searching algorithm of multiresolution window after the pretreatment. After that, the width, amplitude and the difference between the adjacent waveforms of the marked waveforms are selected as the features and the signal is divided into several sections according to the width of marked waveforms. Then the original signal is decomposed by wavelet transform. The ratio of the power of each section to the amplitude of the original one is taken as one feature. Finally, k-means clustering algorithm based on density weighting and preference information is used to recognize and classify the compression and ventilation of the marked waveforms. The experimental results show the accuracy and sensitivity of the recognition are high, the robustness is good and the running time（0.43±0.07 s） can meet the requirement of clinical application.