基于BP神经网络的胎儿心电提取算法研究

Research of fetal electrocardiogram extraction algorithm based on BP neural network

目的针对胎儿心电不易提取的问题,提出一种从孕妇腹部混合心电信号和胸部心电信号中提取胎儿心电的方法。方法采用反向传播(BP)神经网络预测孕妇腹部混合心电信号中母体心电的真实形态,从腹部混合信号中减去预测的母体心电信号便得到胎儿心电信号。与小波阈值去燥算法和自适应滤波算法比较,评价BP神经网络算法可行性。结果相比小波阈值去燥算法和自适应滤波算法,该算法准确度为94.12%,灵敏度为96.97%。这两项指标均优于小波阈值去燥算法的80.52%、93.94%和自适应滤波算法的87.88%、87.88%。结论基于BP神经网络的方法可以提取到纯净的胎儿心电信号,对于胎儿心电监护有一定的应用价值。

Objective To propose one method of extracting fetal electrocardiogram（ECG） from mixed ECG signals of pregnant woman abdomen and her chest ECG in solving the problem of difficult extraction of fetal ECG. Methods The back propaga- tion（BP） neural network was used to predict the real shape of maternal ECG in pregnant woman abdominal mixed ECG signals. The predicted ECG signals were subtracted from abdominal mixed signals to obtain fetal ECG signals. Compared with wavelet threshold denoising algorithm and adaptive filtering algorithm, the feasibility of BP neural network algorithm was evaluated. Results Compared with wavelet threshold denoising method and adaptive filtering algorithm, the algorithm accuracy was 94.12 % and sensitivity was 96.97 %, which were better than those of wavelet threshold denoising method（80.52 %, 93.94 %） and adaptive filtering algorithm（87.88 %, 87.88 %）. Conclusion It is demonstrated that the method based on BP neural net- work could extract nure fetal ECG. which has certain apolication value for fetal ECG monitoring.