Research And Application Of ECG Automated Analysis Algorithm

Cardiovascular is one of the major diseases to threat human life. Theresearch of ECG (electrocardiogram) detection has always attracted theattention of people from different groups, since ECG is the essential way toevaluate cardiac function. ECG shows complex patterns due to thephysiological status of an individual, and ECG varies in accordance withindividual difference. Moreover, the random interference of themeasurement system will deteriorate of the ECG waveform. Exploring ECGfrom the point of view Signal Processing is of the following properties: first,quasi-periodicity signals; second, low-frequency narrow-band signal. Thebiological mechanism of cardiac decides the spectrum of ECG distributes in0.5~50Hz; third, non-stationary signals. Heart beat are influenced byvarious physiological status; fourth, nonlinear time-varying signals. All ofthese lead to certain difficulties in ECG signal detection.ECG detection and analysis is an important part of clinical diagnosis ofheart disease, and the accuracy of it will affect directly the doctor'sjudgment on the patients'condition and the follow-up treatment efficacy.Consequently, how to improve the ECG accuracy, reduce the signal noiseand how to improve the ECG waveform recognition efficiency have beenthe core issues of ECG automatic analysis.On the basis of the ECG data gathered by Chinese Academy ofSciences, Changchun branch, the thesis tries to make an in-depth study onthe automatic analysis algorithm of the standard 12-lead ECG data,including filtering algorithm, ECG waveform recognition algorithm andabnormal ECG analysis algorithm. Furthermore, under the Visual C++programming environment, the thesis designs the system software for12-lead ECG Automatic Analysis on the basis of algorithm analysis. Aiming at the seven common ECG noise, we attempt to pretreatment ECGwith multiple filter technology. Since ECG acquisition systems, acquisitionprocess and the environment are different, the noise type emphasis of ECGwill be different. In order to enhance efficiency and to enhance the filteringeffect of filtering, this thesis introduces the dynamic noise detectiontechnology. That is, it first detects the main type of noise in signals, thendeals with data automatically under the help of corresponding filteraccording to the noise type. This technology makes the five types of filterdesigned for different types of noise more oriented in filtering process. Atthe same time, it is self-adaptive in adjusting and optimizing dynamicallythe parameters of filter on the basis of experiment, so as to further enhancethe ECG filtering effect. In this process, bandpass filter is designed byapplying wavelet transform theory to decompose original ECG intodifferent frequency signals, eliminate some"details", and restore the usefulsignals, in the end, to acquire denoising ECG. This type of filter caneffectively remove baseline drift, Frequency of EMG and interference noise.In the process of ECG waveform recognition, the positioning of QRScomplex is the key and basic part of the analysis. By using wavelettransform technique to extract the features of ECG, we can determine theposition of QRS complex accurately. Based on the analysis of how wavelettransforms on different scale, we can get a more accurate definition forP-wave and T-wave. The basic idea of Error Analysis is introduced on thebasis of the comprehensive judgment of 12-lead ECG. It simplifies QRSintegrated analysis algorithm, improves the accuracy of the positioning ofQRS, and reduce the probability of misjudgment missing. In the basis of theidentification of QRS, we go further to study the positioning of Q-wave,S-wave, the detection of P-wave,T-wave, the extraction of QRS's width andextent, the interval extraction of QT, and the measurement algorithm of ST, Based on the cardiac waveform recognition, this thesis designs anarrhythmia analyzing and processing system. And it also carries out anidentification of 10 typical ventricular arrhythmias with database technology,and the algorithm of ventricular arrhythmia detection. These 10 cardiacdisorders including tachycardia, bradycardia, cardioplegia, Room prematurebeats and arrhythmia, ventricular premature beats sporadic, Paired beat andcoupling, VT and pending wave. The abnormality of QRS wave excludedfrom the 10 typical ventricular arrhythmias is categorized as wave to bedetermined by the doctor in order to avoid missed detections. The resultswill be used to create the 24-hour basis ventricular arrhythmias report; HRtrend and ST segment trend diagrams.Some of the software mentioned in this paper is carried out in theVisual C++ environment. The analytical and controlling software providesthe controlling interface for the data acquisition box and also creates themanagement program for the user's database which integrates the collection,analysis and management of the electrocardiosignal. A report of 24-hourbasis arrhythmia, HR trends and ST-segment trends diagrams will becreated by a very easy procedure. In addition, other practical functions suchas the transmission, storage of the electro cardio data, the management ofthe electro cardio data, electrocardiosignal wave browse, printing, storage ofthe analytical results and output are also assets.This thesis carries out actual tests on the system detection analysissoftware with the clinical ECG database and 13 typical cases. The resultindicates that the 12-lead ECG analysis software has reached to the standardand met the basic needs of the clinical diagnosis, which can serve as avaluable resource to the diagnosis of cardiovascular disease.