摘要
目的研究一种基于多心电(ECG)周期融合和优先权分类的心室早期收缩(premature ventricular contraction,PVC)高精度检测方法。方法利用再定义ECG样本和2种不同ECG分割方法得到4个以非线性Hermite系数为特征的向量集。文中的数据取自MIT-BIH数据库,包括正常窦性心律(normal sinus rhythm,NSR)和PVC。进行一种基于类优先条件约束的改建二次判别函数(improved quadratic discriminant function,IQDF)的分类,其中以贝叶斯分类阈值为基准寻找在优先限定PVC错误率条件下使NSR错误率为最小的拉格朗日分类阈值。结果 PVC和NSR分别取得了99.29%和96.73%的分类精度。结论文中方法不仅能使PVC高分类精度得到优先保证,而且能使NSR分类精度保持在理想的高水平上。
Objective To study high accuracy method for detecting premature ventricular contraction (PVC) based on multiple cardiac cycle fusion and prior classification. Methods Four different feature vector sets of nonlinear Hermite coefficient features were obtained with redefinition of electrocardiogram (ECG) samples and 2 different ECG segment methods. The data for this paper were taken from MIT-BIH database, including PVC and normal sinus rhythm (NSR). The classification, based on an improved quadratic discriminant function (IQDF) constrained by prior-like condition, was carried out. During classification, Lagrange classification threshold was determined at a fiducial point of Bayes classification threshold in order to find out the minimal NSR classification error under prior PVC misclassification rate. Results Experimental results showed that the accuracy of 99.29% and 96.73% were achieved for detecting PVC and NSR respectively. Conclusion This proposed technique not only can have a prior to high-accuracy classification for PVC, but also can keep classi- fication accuracy in high level as soon as possible for NSR.
出处
《航天医学与医学工程》
CAS
CSCD
北大核心
2014年第1期20-25,共6页
Space Medicine & Medical Engineering
基金
浙江省自然科学基金资助课题(Y1100219)
关键词
心室早期收缩
多ECG周期
特征提取
贝叶斯
优先分类
premature ventricular contraction
multiple ECG cycles
feature extraction
Bayes
prior classifica-tion