中文摘要：

左束支传导阻滞（LBBB）作为临床常见的一种心律失常,是左心室收缩功能减低、患者死亡率增加的标志;利用机器学习算法对其进行辅助诊断,将对LBBB早发现、早治疗起到积极的推动作用。然而,由于目前常用的支持向量机（SVM）等传统的机器学习算法容易产生局部最优解,准确度有待提高,因此提出一种基于极限学习机（ELM）的LBBB辅助诊断算法。首先,利用小波进行心电信号预处理,包括基线漂移、肌电噪声及工频干扰的去除;接着,确定QRS波群与T波位置;然后,根据临床上LBBB患者比正常人的QRS波群持续时间延长等特点,建立融合时域、形态与能量3类特征的特征模型;最后,利用该模型提取的特征集合,提出基于ELM的LBBB辅助诊断算法。此外,在MIT_BIH数据库中的5 000份ECG数据上进行实验验证,结果表明所提出的预处理与波形提取算法能有效去除噪声并提取QRS-T特征波;在LBBB的判别上,相比SVM算法、ELM算法的训练时间缩短了88.5%;同时,在准确率、灵敏度、特异度、LBBB检出率和正常人检出率的指标上,分别提升2.4%、5.4%、1.2%、3.6%和2%。因此,基于ELM的LBBB辅助诊断算法具有明显优势。

英文摘要：

As a common clinical arrhythmia, left bundle branch block is a signal of left ventricular systolic function decreased and mortality increased in patients, machine learning algorithm aided diagnosis of the disease will play a positive role in detection and diagnosis. Currently, left bundle branch block automatic identification mode is still using support vector machines and other traditional machine learning algorithms for training and testing, these traditional neural network algorithms prone to local optimal solution, which is not suitable to classified LBBB. Herein, this paper proposed an algorithm about automatic diagnosis of left bundle branch block based on ELM. Firstly, the ECG signal was preprocessed, including the removal of baseline drift, high-frequency noise and power-line interference; then, we created the model by features of LBBB such as the length of QRS after the location of QRS-T wave was determined. Finally, we provided the LBBB diagnosis algorithm based on ELM. Additionally, we tested 5000 groups of data in MIT_BIH. Results showed the algorithm was effective in noise removal and wave extraction. ELM was 88.5% that is shorter than SVM in training time, and ELM had improvement of 2. 4%, 5.4%, 1.2%, 3.6%, 2% in time, accuracy, sensitivity, specificity, FP ratio and FN ratio respectively. Accordingly, ELM had more advantages in LBBB diagnosis.