Yang Juanjuan, Xi Caiping
Ocean College, Jiangsu University of Science and Technology, Zhenjiang 212100, China.
College of Automation, Jiangsu University of Science and Technology, Zhenjiang 212100, China.
Entropy (Basel). 2022 Dec 1;24(12):1763. doi: 10.3390/e24121763.
Congestive heart failure (CHF) is a chronic heart condition associated with debilitating symptoms that can lead to mortality. The electrocardiogram (ECG) is a noninvasive and simple diagnostic method that can show detectable changes in CHF. However, manual diagnosis of ECG signals is often erroneous due to the small amplitude and duration of the ECG signals. This paper presents a CHF diagnosis method based on generalized multiscale entropy (MSE)-wavelet leaders (WL) and extreme learning machine (ELM). Firstly, ECG signals from normal sinus rhythm (NSR) and congestive heart failure (CHF) patients are pre-processed. Then, parameters such as segmentation time and scale factor are chosen, and the multifractal spectrum features and number of ELM hidden layer nodes are determined. Two different data sets (A, B) were used for training and testing. In both sets, the balanced data set (B) had the highest accuracy of 99.72%, precision, sensitivity, specificity, and F1 score of 99.46%, 100%, 99.44%, and 99.73%, respectively. The unbalanced data set (A) attained an accuracy of 99.56%, precision of 99.44%, sensitivity of 99.81%, specificity of 99.17%, and F1 score of 99.62%. Finally, increasing the number of ECG segments and different algorithms validated the probability of detection of the unbalanced data set. The results indicate that our proposed method requires a lower number of ECG segments and does not require the detection of R waves. Moreover, the method can improve the probability of detection of unbalanced data sets and provide diagnostic assistance to cardiologists by providing a more objective and faster interpretation of ECG signals.
充血性心力衰竭（CHF）是一种慢性心脏疾病，伴有使人衰弱的症状，可导致死亡。心电图（ECG）是一种无创且简单的诊断方法，能够显示CHF中可检测到的变化。然而，由于ECG信号的幅度小和持续时间短，手动诊断ECG信号往往有误。本文提出了一种基于广义多尺度熵（MSE）-小波首波（WL）和极限学习机（ELM）的CHF诊断方法。首先，对来自正常窦性心律（NSR）和充血性心力衰竭（CHF）患者的ECG信号进行预处理。然后，选择分割时间和尺度因子等参数，并确定多重分形谱特征和ELM隐藏层节点数量。使用两个不同的数据集（A、B）进行训练和测试。在这两个数据集中，平衡数据集（B）的准确率最高，为99.72%，精确率、灵敏度、特异性和F1分数分别为99.46%、100%、99.44%和99.73%。不平衡数据集（A）的准确率为99.56%，精确率为99.44%，灵敏度为99.81%，特异性为99.17%，F1分数为99.62%。最后，增加ECG段数量和不同算法验证了不平衡数据集的检测概率。结果表明，我们提出的方法所需的ECG段数量较少，且不需要检测R波。此外，该方法可以提高不平衡数据集的检测概率，并通过对ECG信号提供更客观、更快的解释为心脏病专家提供诊断帮助。
