基于稀疏主元分析的微伏级T波交替幅度量化

Quantifying the Amplitude of Microvolt T Wave Alternans with Sparse Principal Component Analysis

T波交替是心电(ECG)信号中重要的猝死预测因子,其幅度量化具有重要的临床意义。然而,T波交替处于微伏级,并且具有非平稳性,导致精确量化其幅度存在较大困难。本文以逐搏ECG信号的主元分析(PCA)为基础,运用统计检验方法挑选包含T波交替的主元。通过引入稀疏性约束,排除具有较小变异性的变量,实现了主元选择和变量选择。以国际标准数据库的ECG信号为例,在等级相关系数和相对误差两项指标上,对比多种量化方法,本文提出的基于稀疏主元(SPC)的量化方法均具有优势。

T wave alternans in ECG is an important prediction factor for a sudden death. It is crucial in clinic to quan tify the amplitude of T wave alternans. However, T wave alternans are highly non-stationary, which makes accurate quantification very difficult. In this study, we proposed a new method to improve the amplitude quantification. We identified T wave alternans by principal component analysis （PCA） with statistical test, and processed the principal components with T wave alternans further by imposing sparse constraint. We evaluated and compared our method a- gainst other 4 popular solutions on an international benchmark database. The results including rank correlation coef- ficient and relative error indicate that this sparse principal component （SPC） based method is better than others.