基于极限学习机的室性早搏判别算法的实现

发布时间：2018-05-14 10:33

本文选题：极限学习机 + 室性早搏　；参考：《中国生物医学工程学报》2017年02期

【摘要】：室性早搏是常见的心律异常疾病,给人的生命带来威胁,准确的心律异常诊断对于帮助人们预防心血管疾病起到重要的作用。以MIT-BIH心律异常数据库中的数据作为分析对象,提出一种基于极限学习机算法的诊断方法,主要包括信号预处理、特征提取和分类,实现心电信号室性早搏异常的判别。采用小波变换结合形态学算法,对信号进行预处理,去除干扰,得到纯净的心电信号。通过K-means聚类算法提取QRS波群等特征参数,根据这些参数建立正常窦性心律和室性早搏的正样本和预测样本,再结合极限学习机分类器进行样本训练匹配和分类识别。选取1 260个周期信号进行实验,结果表明,该算法能准确诊断出室性早搏异常,最终阳性平均检测率达到95%,平均灵敏度达到96%。该算法相比其他算法,在识别精度相当的情况下,可极大提高算法的实时性,具有很高的研究价值,同时在移动医疗和临床医疗方面也具有一定的实用价值。
[Abstract]:Ventricular premature beat (VPB) is a common arrhythmia disease, which is a threat to human life. Accurate diagnosis of ventricular premature beat plays an important role in the prevention of cardiovascular disease. Taking the data of MIT-BIH arrhythmia database as the analysis object, a diagnosis method based on extreme learning machine algorithm is proposed, which mainly includes signal preprocessing, feature extraction and classification, so as to distinguish ventricular premature beat abnormality of ECG signal. Wavelet transform combined with morphological algorithm is used to preprocess the signal to remove interference and get pure ECG signal. The characteristic parameters of QRS wave group were extracted by K-means clustering algorithm. According to these parameters, positive and predictive samples of normal sinus rhythm and ventricular premature beats were established, and then the training matching and classification recognition of samples were carried out by combining with the extreme learning machine classifier. The experimental results of 1 260 periodic signals show that the algorithm can accurately diagnose abnormal ventricular premature beats, and the average positive detection rate is 95% and the average sensitivity is 96%. Compared with other algorithms, this algorithm can greatly improve the real-time performance of the algorithm under the condition of the same recognition accuracy, and has high research value, and also has certain practical value in mobile and clinical medicine.
【作者单位】：杭州电子科技大学生命信息与仪器工程学院;杭州红十字会医院骨科;杭州红十字会医院医务科;
【基金】：国家自然科学基金(61374005)
【分类号】：R541.7;TN911.7

【相似文献】