摘要
建立了基于心率变异性(heart rate variability,HRV)信号分析的人工神经网络(artificial neural network,ANN)模型,以寻求用于充血性心衰(congestive heart failure,CHF)诊断的最佳向量-网络组合。结果表明,将经过改进的BP算法和小波分析所抽取的特征向量相结合所获得的神经网络在诊断敏感性和特异性上有着均衡且优良的表现,并且经由AR模型谱估计获取的向量价值也不亚于小波分析所提取的特征向量。因此,基于HRV信号分析的人工神经网络用于诊断CHF可作为临床诊断的一种重要参考方法。
In this paper we build Artificial Neural Network（ANN） Models which is based on heart rate variability （HRV） analysis, for seeking the best combination of input vectors and ANN for congestive heart failure（CHF） diagnosis. It is proved that the ANN which combining the improved BP algorithm and feature vectors extracted by using wavelet analysis, presents great and balanced performance in sensitivity and specificity. And the value of feature vectors extracted by AR spectrum estimation model may be not less than that by wavelet analysis. Consequently diagnoses on CHF by using ANN based on HRV analysis, should be a important reference approach for clinical diagnoses.
出处
《北京生物医学工程》
2007年第2期182-186,共5页
Beijing Biomedical Engineering
