Localization of Ventricular Activation Origin from the 12-Lead ECG: A Comparison of Linear Regression with Non-Linear Methods of Machine Learning
Localization of Ventricular Activation Origin from the 12-Lead ECG: A Comparison of Linear Regression with Non-Linear Methods of Machine Learning
复制标题
DOI：
10.1007/s10439-018-02168-y
复制
发表时间：
2019-02-01
影响因子：
3.8
通讯作者：
Horacek, B. Milan
中科院分区：
文献类型：
--
作者：
Zhou, Shijie;AbdelWahab, Amir;Horacek, B. Milan
我们已经开发了一种基于多元线性回归(MLR)模型的自动定位方法，可以从12导联心电图实时估计普通左心室(LV)心内膜表面的激活起源。本研究旨在探讨机器学习，即随机森林回归(RFR)和支持向量回归(SVR)是否能比MLR提高定位精度。38名患者在1012个部位的左心内膜起搏时获得12导联心电，已知坐标从电解剖标测系统输出；然后将每个起搏部位登记到被细分成238个三角形的16个节段的普通左室心内膜表面。心电图减少到每个导联一个变量，由QRS的120ms时间积分组成。为了比较三种回归模型，将整个数据集(n=1012)随机分成80%的设计集和剩余20%的测试集，并用普通LV表面上的测地线距离来评估定位误差。带替换的Bootstrap方法，使用1000次重抽样试验，估计了每个模型对遗漏样本(n类似或等于371)的误差分布。在设计集(n=810)中，SVR、RVR和MLR的平均精度分别为8.8、12.1和12.9 mm。在测试集(n=202)中，SVR模型的定位误差的平均值始终低于其他两个模型，均低于MLR(11.4比12.5 mm)和RFR(11.4比12.0 mm)；在估计定位精度方面，RFR模型也优于MLR模型(12.0比12.5 mm)。1,000个样本的Bootstrap方法证实，SVR和RFR模型在样本遗漏的Bootstrap评估中的预测准确率显著高于MLR(SVR与MLR(p
We have previously developed an automated localization method based on multiple linear regression (MLR) model to estimate the activation origin on a generic left-ventricular (LV) endocardial surface in real time from the 12-lead ECG. The present study sought to investigate whether machine learningnamely, random-forest regression (RFR) and support-vector regression (SVR)can improve the localization accuracy compared to MLR. For 38 patients the 12-lead ECG was acquired during LV endocardial pacing at 1012 sites with known coordinates exported from an electroanatomic mapping system; each pacing site was then registered to a generic LV endocardial surface subdivided into 16 segments tessellated into 238 triangles. ECGs were reduced to one variable per lead, consisting of 120-ms time integral of the QRS. To compare three regression models, the entire dataset (n=1012) was partitioned at random into a design set with 80% and a test set with the remaining 20% of the entire set, and the localization errormeasured as geodesic distance on the generic LV surfacewas assessed. Bootstrap method with replacement, using 1000 resampling trials, estimated each model's error distribution for the left-out sample (n similar or equal to 371). In the design set (n=810), the mean accuracy was 8.8, 12.1, and 12.9mm, respectively for SVR, RVR and MLR. In the test set (n=202), the mean value of the localization error in the SVR model was consistently lower than the other two models, both in comparison with the MLR (11.4 vs. 12.5mm), and with the RFR (11.4 vs. 12.0mm); the RFR model was also better than the MLR model for estimating localization accuracy (12.0 vs. 12.5mm). The bootstrap method with 1,000 trials confirmed that the SVR and RFR models had significantly higher predictive accurate than the MLR in the bootstrap assessment with the left-out sample (SVR vs. MLR (p