AMBULATORY MONITORING AND ANALYSIS OF ECG SIGNALS

Abstract

Ambulatory monitoring of ECG signals is an important topic of research in the area of Biomedical Engineering. The basic concept lies in providing medical service to the cardiac patients away from the hospital. A complete computer based management ofthe ambulatory monitoring ofECG forms the basis of the present work. Two improtant aspects, namely devel opment of ambulatory monitoring system for arrhythmia analysis and data compression techniques for both, on-line and off-line, applications were identified for the development of this concept by the author. A good number of ambulatory monitoring systems are available for arrhythmia analysis ofambulatory patients. Various techniques and algorithms for arrhythmia analysis and data compression are available. More important once are briefly reviewed in the thesis. A prototype ambulatory monitor has been developed based on 8088 microprocessor system. There are five modules in the ambulatory monitor ing unit. The first one is the data acquisition module capable of recording three channels simultaneously. The input ECG signal is conditioned using instrumentation amplifier and highpass, notch and lowpass filters in cascade for eliminating baseline wander, power line interference and high frequency noise contents, respectively. The second module consists of 8088 micro processor and accessories, including EPROM which is loaded with the software.The module acts as the controller of the ambulatory monitor. The digital input/output module consist of different programmable peripheral chips for providing timing signals, parallel to serial data communication and pushbutton / keyboard controls. The memory module consists of three banks of read/write memories , each bank providing 256 k bytes for ECG data storage. The indication and display module gives a visual indication of different events and alarms and a numerical display of the crrent values of ECG parameters, such as heart rate, intervals etc. All the five modules inter faced to each other through a mother board are controlled by a sophisti cated and efficient software package developed using assembly level programming . Having developed the hardware for ambulatory monitor, two important algorithms were developed for processing the ECG signal. The first algo rithm is for eleminating various types of noise present in the ECG signal, but not completely eleminated by the hardware . The second algorithm developed provides ECG wave recognition to measure amplitudes (P, Q, R, S and T waves) and the characteristic points (Pon , Poff, QRSon, QRSoff and Tend) of the ECG. The detection of QRS complex has been obtained through slope-threshold method. The baseline estimation is made according to the recommendations of the Common Standards for Quantitative Electro cardiography (CSE) Working Party. The parameters obtained through this software are stored in a file for further processing and use. • For arrhythmia analysis, the commonly used ECG features are intervals and segments; these are derived using the parameters obtained from the software developed for the ECG wave recognition and characteristic point location. Diagnostic criteria are then applied for classifying the arrhythmia. The five different types of arrhythmias for which diagnostic rules are clearly known namely, Normal sinus rhythm, Atrial premature contraction, Sinus arrhythmia, First degree AV block and Sinus bradycardia, are identified from the derived features of ECG based on these rules. A six - step logical procedure has been adopted in the algorithm for arrhythmia classification. The algorithm thus developed has been tested on different records of Massa chusetts Institute of Technology/Beth Israle Hospital (MIT/B1H) arrhythmia database. These results are then compared on a beat-by-beat basis, each beat being individually examined and verified against the results available in the database. Another aspect of the thesis is data compression, transmission and reconstruction of the original signal. Data compression becomes neces sary to reduce the memory requirement for storage and to speed-up data transmission in real-time. Various direct data compression techniques exist in the literature; five out of these techniques have been selected for compari son and evaluation. The performance of these techniques have been mea sured using two indices, namely, compression ratio and percent root-meansquare difference besides fidelity of the reconstructed signal. The major contribution in this work is the study of the effect of sampling frequency on the performance of direct data compression techniques and the clinical acceplabilty of such schemes. In order to know the clinical acceptable qual ity of the reconstructed signal, peak, boundary and interval measurements were made both on the reconstructed and the original signals of the sanie record and the results were compared. i/ Data compression techniques have also been proposed for offline stor age as database. Two transform based methods namely, fast Fourier trans form (FFT) and fast Walsh transform (FWT) based on the principle of suc cessive doubling, have been developed for the purpose. The performance of these compression schemes are evaluated using compression ratio and PRD ,besides fidelity of the reconstructed signal. Further, to know the extent to which the clinical information is preserved in the reconstructed signal, peak, boundary and interwave measurements were made on both the original and reconstructed signals and compared. The algorithms for real-time implementation in ambulatory monitor have been selected on the basis of simplicity and speed of execution to implement on microprocessor. Algorithms for data acquisition, data com pression, QRS detection and arrhythmia analysis have been developed and implemented using assembly language. The performance of the algo rithms have been tested individually. The data acquisition algorithm has been tested by acquiring the ECG signal from the normal human beings. The ECG recorded is of good quality and can be used straightaway for further analy sis. Other algorithms namely data compression, QRS detection and arrhythmia interpretation have been tested on standard MIT/BIH Arrhythmia data base. The performance of all these algorithms is found to be satisfactory. The results obtained from different algorithms tested on the stan dard CSE & MIT/BIH database are critically examined in the thesis. The conclusions in respect of the work presented are given at the end of the thesis and finally the scope for future improvements is brought out.