心电信号在线数据知识化辅助诊断算法研究

Knowledge discovery based auxiliary diagnosis algorithm for electrocardiograph on-line data

针对心电监护与诊断过程中数据量大、准确性和快速性要求高的特点,提出了一套基于数据知识化的心电辅助诊断算法.该套算法包括数据识别、冗余处理、转换和提取过程,利用小波变换的多分辨率和抗干扰能力好的特点,检测QRS波、P波、T波,提高了特征检测的准确性;利用聚类分析具有较好的鲁棒性和适合于大数据量分析的特点,对QRS进行波形分类;算法结合了单独一搏诊断和串诊断以及多参数综合分析.采用MIT-BIH标准心电数据库中的部分数据和心电专家确诊的心律失常数据文件对该算法进行了评估,检出率都在95%以上,表明该套算法对部分心律失常可以进行有效分析.

A series of data knowledge discovery based electrocardiograph （ECG） auxiliary diagnosis algorithms were presented against the characteristics of huge data quantum, high accuracy and rapidity demands in the ECG monitor and diagnosis process. The algorithm consists of several stages, including data distinguishing, data redundant processing, data conversion and data extraction. The characteristics of wavelet transform, multiresolution and high anti-interference, were used to detect QRS, P and T waves and improve the accuracy of character detection. Clustering analysis characterized by better robustness and capability to analyze huge data quantum was used to classify QRS wave. The algorithm combines diagnosis based on one beat, string diagnosis and comprehensive analysis with multiparameters. Verified by partial data of MIT-BIH standard ECG database and arrhythmia data files diagnosed by ECG experts, the detectability exceeded 95 %, which showed that the algorithm could analyze partial arrhythmias effectively.