Tap The Singular Structure Of The Ecg Time Series

Cardiovascular diseases have now become major killers for humans. The ECG time series is one of the main means to diagnose cardiovascular diseases. The development of the methods and technology for measurement in the modern medical field has made it possible to record large amount of time series. To realize efficacious analysis using the enormous amount of medical information, a high performance and automatic data-analysis technique is required.After the inventation of ECG one hundred years ago, the main methods to analyze and manage the ECG time series are based on certain settled templets or the comparison of similarity between different time series. Investigators extracted various typical wave types, used them as the templets for the analysis of ECGs, and had achieved good results. Unfortunately, this method with the basis on templets or similarity comparison has its own limitation. Using this method, if a particular wave structure does not match any templet in the existing templet databases, it is then hardly to be recognized, which might lead to mistakes in the diagnosis and prognosis of the diseases. In this study, we tried to locate the abnormal waveforms in ECG from a different angle. Instead of matching a templet, we tried to find the waveform which was the most dissimilar to the other waveforms. In this study, we named this waveform or subsequence of the ECG time series a discord.Although there were relatively accurate algorithms for finding discords in the time series, they are hardly to be used, because that the algorithms was resource and time consuming. In our research, the algorithms of discords searching was optimized so that it could run on a regular PC and could be accessed from a web browser. In our project, we chose the Eclipse + Tomcat integrated development environment (IDE), used the Java + Matlab mixed programming technology, the B/S mode and classical MVC architecture, and finally developed an experimental web application system for the sequence minning of discords in the ECG time series. The system is composed by 7 modules: analytic module, ECG denoising module, standardization module, configuration of the best size of the sliding window, SAX module, discords searching module and interaction module. The whole platform consists of over 50 files and around 10,000 code lines. Our research alsobuilt the ECG database for the application system, which can be used through the web application system. The main data resources are MTT/BIH standard ECG data, the ECG data of PUMC Hospital.The user can access the system through a web browser and process the data chosen from the databases. The discords in the ECG time series can be located promptly and accurately. The results could be displayed in the browser. The discords found by this system were consistent with these remarked in MIT/BIH database.The discord searching in ECG time series is a new technique. This study was only primary experiments on the feasibilities of the algorithms and experiment platform, and seta basis for further researches. The clinical application of the discords in the ECG time series deserve further investigations.