便携心电传感器的心房肥大诊断研究

基于便携式传感器的模式识别在心电(ECG)监护领域具有广泛的应用前景,并且在心律不齐、心肌梗塞、心室肥大等心电的识别算法上都已有大量的研究与应用,但在心房肥大诊断上却未有模式识别相关的研究成果。心房肥大病症的心电数据量不足给研究造成重大障碍,部分分类器无法适应小样本训练下的分类。针对小样本训练进行研究,对比了不同分类方法,显示了基于统计模式识别的支持向量机(SVM)应用于心房肥大的应用潜力。另外,由于不同个体的心房肥大心电存在差异,在实际应用环境中,SVM存在无法良好泛化的问题,存在类别错分的医学风险。针对类别错分情况,采用分类器融合的方法改进分类器,提出了在SVM分类器输出端增加了拒绝域的分类器(SVM-R)的方法。实验结果表明:SVMR有较高的分类准确率与诊断可信度。

Research on atrial hypertrophy diagnosis by portable ECG sensor

Pattern recognition method based on portable sensor has broad application prospects in electrocardiograph（ECG） diagnostic areas, there are a lot of research and application in arrhythmia, myocardial infarction, ventricular hypertrophy recognition, but there is no relevant research about pattern recognition method in atrial cardiac hypertrophy. Atrial hypertrophy ECG data is insufficient, which cause significant obstacles for research, and part of classifier can ＇ t adapt to the classification of small sample case. Studies of small sample training is focused on, compare different classification methods, it shows great application potential of support vector machine （ SVM ） based on statistical pattern recognition in atrial hypertrophy. In addition, as different individuals, atrial cardiac hypertrophy is different, in actual application environment, SVM has generalization problem ,and there is medical risk of category misclassified. Aiming at category misclassification cases, using classifier fusion method to improve classifier, adding rejection region classifier at the output end of SVM classifier （SVM-R） is proposed. Experimental results show that SVM-R has higher classification accuracy and diagnostic corffidence.