The Key Lab for Health Informatics, Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences, Shenzhen, China.
Comput Biol Med. 2013 Mar;43(3):176-83. doi: 10.1016/j.compbiomed.2012.12.005. Epub 2013 Jan 11.
Two different methods for extracting atrial activity (AA) signal from single lead electrocardiogram (ECG) of atrial fibrillation were proposed. The first one is a weighted average beat subtraction (WABS) method. Coefficients of QRS complexes used for constructing QRS template were obtained by minimizing mean square error. The second method is based on maximum likelihood estimation (MLE). Probability density functions of AA signal and ventricular activity (VA) signals were estimated using generalized Gaussian model. Then AA signal was extracted by maximizing likelihood function. Simulated signal and clinical ECG were used to evaluate the performance of ABS, WABS and MLE-based algorithm. In comparison with ABS, WABS and MLE-based algorithm reduced normal mean square error by 23.5% and 20.2%, respectively.
提出了两种从心房颤动单导联心电图中提取心房活动（AA）信号的不同方法。第一种是加权平均节拍减法（WABS）方法。用于构建 QRS 模板的 QRS 复合系数是通过最小化均方误差获得的。第二种方法基于最大似然估计（MLE）。使用广义高斯模型估计 AA 信号和心室活动（VA）信号的概率密度函数。然后通过最大化似然函数来提取 AA 信号。使用模拟信号和临床 ECG 来评估 ABS、WABS 和基于 MLE 的算法的性能。与 ABS 相比，WABS 和基于 MLE 的算法分别将正常均方误差降低了 23.5%和 20.2%。
