基于多指标融合的动态心电质量评估

Quality Assessment of Dynamic Electrocardiogram Based on the Fusion of Multiple Parameters

动态心电信号易受环境和行为等因素的干扰,其质量难以保证。本文在对现有心电质量评估技术进行深入研究的基础上,提出了基于多指标融合的动态心电质量评估方法。该方法主要由指标提取、融合与综合评估三部分组成。指标分为外层指标与内层指标,前者依据电极片脱落检测指标,将信号分为可用与不可用;后者依据时域、频域分析的结果,分别提出基于小波变换的基线漂移程度指标、基于上下包络差的幅值变化程度指标以及基于功率谱估计的频率成分指标。对于融合部分,内层指标融合采用信息熵加权的加法合成法,内层与外层之间融合采用乘法合成法。综合评估采用区间估计的方法得到阈值,进一步划分信号质量的好与坏。通过比较两种方法,结果表明区间估计法(准确度:96.24%;灵敏度:97.83%)明显优于经验阈值法(准确度:92.39%;灵敏度:95.16%)。本文提出的评估方法是可行的,为动态心电信号提供了一种更加精确的质量评估方法。

Dynamic electrocardiogram （ECG） signal is easily affected by environmental and behavioral factors. Based on the research of ECG quality assessment technologies, this paper proposed a quality assessment method on the basis of multi-index fusion, which is composed of three parts： index extraction, fusion and quality assessment. Indices can be divided into outer layer index and inner layer index. The former one is electrode fall-off detection index proposed by electrode fall-off experiment, while the latter is extracted from analysis in both time domain and frequency domain, including baseline wander level index based on wavelet transform, amplitude change level index based on the difference between top and bottom envelopes transform, and the frequency component index based on power spectrum estimation. As to fusion, this paper conducted additive synthesis based on entropy weight to the fusion of inner layer indices, and multiplicative synthesis to the fusion of inner and outer layer indices. Comprehensive assessment classifed signals into good and bad using the threshold acquired from the interval estimation method. The results showed that the interval estimation method （accuracy： 96.24%; sensitivity： 97.83%） was signifcantly better in comparison with the empirical threshold method （accuracy： 92.39%; sensitivity： 95.16%）. The proposed method is feasible, and provides a more accurate evaluation on the quality of dynamic ECG signals.