基于优化自适应模型的心律失常辅助诊断方法

An Auxiliary Diagnosis Method for Arrhythmias Based on Optimized Adaptive Model

针对心律失常诊断算法中存在的不平衡数据集诊断准确率及阳性预测值较低的问题,提出一种基于优化自适应模型的心律失常辅助诊断方法;提取心电信号的77维特征并将其融合,使用融合特征训练诊断模型,同时利用改进的粒子群算法优化自适应模型参数;采用优化模型对MIT-BIH心律失常数据库进行诊断实验并与现有方法进行对比。结果表明,本文所提方法在测试数据集的诊断准确率达到98.2%,正常或束支传导阻滞节拍、室上性异常节拍、心室异常节拍、融合节拍的阳性预测值分别达到98.5%、96.1%、95.5%、92.0%,诊断准确率和阳性预测值明显大于现有方法的。

Aiming at the problem of low diagnostic accuracy and positive predictive value of imbalanced datasets in arrhythmia diagnosis algorithms,an auxiliary method for arrhythmia diagnosis based on optimized adaptive models was proposed.This method extracted the 77 dimensional features of the electrocardiogram signal and fuses them,trained the diagnostic model using the fused features,and optimized the adaptive model parameters using an improved particle swarm optimization algorithm.The optimized model was used to test in MIT-BIH arrhythmia database and compared with the existing methods.The results show that the total diagnostic accuracy of the proposed method on the test dataset reaches 98.2%,and the positive predictive values of normal or bundle branch block rhythm,supraventricular abnormal rhythm,ventricular abnormal rhythm,and fusion rhythm reach 98.5%,96.1%,95.5%,and 92.0%,respectively.The diagnostic accuracy and positive predictive value are significantly higher than those of the existing methods.