The main purpose of the present work is to design and implement a prototype ECG system with wireless links for continuous monitoring of the subject for cardiac related problems. The ECG signal acquired from subject is filtered, digitized, and compressed for wireless communication. The proposed system can be extended, upon interfacing with other devices, for continuous monitoring of other vital parameters of the patient.
In automation of the ECG signal analysis, the workload of the medical professionals can be reduced. The automated system provides an alert when critical changes are detected by the system. Concisely stated, ECG of the patient is continuously monitored and deviations from normalcy are detected in real-time. The changes in the ECG could
be due to heart attack, fibrillation or arrhythmias. In case of emergency, data is transmitted to a medical practitioner, who in turn can provide necessary directions to take care of the situation. In this manner, as the problems can be detected as and when they occur, the remedial actions are initiated before the problems become serious.
The complete ECG diagnostic system includes a low power Instrumentation amplifier, filters, ADC, Microcontroller and ZIGBEE modules. MATLAB / LABVIEW are used for signal analysis and classification. These environments are capable of not only collecting, recording, transmitting, and displaying ECG data on a real time basis but also for analyzing the acquired ECG data in order to detect the cardiac abnormalities.
The MIT-BIH database signals were used for validation and evaluation of classification algorithms. In order to reduce the memory requirements for storing the acquired ECG signals, ECG data was compressed. Discrete Cosine Transform (DCT) technique was applied for ECG data compression. Here DCT showed good performance with a Compression Ratio (CR) of 82-90.43% and Percent Root Mean Difference (PRD) of 7.9-0.93. Linear Vector Quantization method (LVQ)
is used for identifying the abnormalities associated with the ECG signal. After training the LVQ process with a reasonable number of samples, the algorithm is used for classifying ECG signals. The techniques used in the present work for ECG signal compression and classification gave better results compared to those found in the literature.
CONTENTS
DECLARATION
CERTIFICATES
ACKNOWLEDGEMENTS
ABSTRACT
CONTENTS
LIST OF TABLES
LIST OF FIGURES
NOTATIONS
CHAPTER 1 INTRODUCTION
1.1 ECG and its Importance
1.2 ECG Lead System
1.3 Need for ECG Monitoring System
1.4 Literature Survey
1.5 Methodology
1.6 Organization of thesis
CHAPTER 2 ECG DATA ACQUISITION SYSTEM AND HEART RATE MEASUREMENT
2.1 ECG System Requirements
2.1.1 Data Acquisition Unit
2.1.2 Data Processing Unit
2.1.3 Data Communication Unit
2.1.4 Data Analysis Unit
2.2 ECG Signal Data Acquisition
2.2.1 Electrodes used for ECG Signal Pickup
2.2.1.1 Surface Electrodes
2.2.1.2 Adhesive Electrodes
2.3 Instrumentation Amplifier
2.3.1 Requirements of Instrumentation Amplifier
2.3.2 AD620 Instrumentation Amplifier
2.4 Simple ECG Acquisition System
2.4.1 Filters Used in ECG System
2.4.1.1 Low Pass Filter (LPF)
2.4.1.2 High Pass Filter (HPF)
2.4.1.3 Notch Filter
2.5 Data Processing
2.5.1 ADC0804 Analog to Digital Converter
2.5.2 Serial communication using microcontroller 89C
2.6 Power Supply
2.7 Data Communication Unit
2.7.1 Communication of Data HyperTerminal
2.7.1.1 Steps involved to setup a new connection using window interface
2.7.1.2 Steps involved in saving incoming data to a text file
2.8 Realization of Sigma Delta ADC using Simulink
2.9 Heart Rate Measurement using LABVIEW
2.9.1 Implementation of LABVIEW for ECG Instrumentation and Analysis
2.9.2 Data Acquisition Module
2.9.3 Amplification Module
2.9.4 Filtering module
2.9.5 QRS Detection and Heart Rate Calculation
CHAPTER 3 ECG COMPRESSION TECHNIQUES
3.1 Performance Evaluation of compression
3.1.1 Compression Measurement
3.1.2 Distortion Measurement
3.2 Data Compression
3.2.1 Direct Data Compression
3.2.2 Transformation Methods
3.3 Compression techniques used in the Proposed Work
3.3.1 Amplitude Zone Time Epoch Coding (AZTEC) Algorithm
3.3.1.1 Line Detection (Horizontal Mode) Procedure
3.3.1.2 Line Processing (Slope Mode) Procedure
3.3.2 Turning Point (TP) Algorithm
3.3.3 Coordinate Reduction Time Encoding System (CORTES) Algorithm
3.3.4 Discrete Cosine Transform (DCT) Algorithm
3.3.5 Fast Fourier Transform (FFT) Algorithm
3.4 Comparison of various Compression Techniques
3.4.1 Conclusions
CHAPTER 4 LINEAR VECTOR QUANTIZATION FOR ECG SIGNAL CLASSIFICATION
4.1 Feature Extraction
4.1.1 QRS Detection
4.1.2 R-R Interval Calculation
4.1.3 ST Segment Measurement
4.1.4 Heart Rate Determination
4.2 Artificial Neural Network
4.2.1 Training the Neural Network
4.2.2 Supervised Training for Neural Network
4.3 Linear Vector Quantization (LVQ)
4.3.1 ECG Signal Data Set
4.3.2 Training Algorithm for ECG Signal Classification
4.3.3 Application of the ECG Signal Analysis using LVQ Method
4.3.4 Classifier Performance
CHAPTER 5 RESULTS OF ECG ACQUISITION, COMPRESSION AND ANALYSIS
5.1 Hardware Implementation for Data Acquisition
5.2 Results of the ECG Compression
5.3 Results of Linear Vector Quantization (LVQ) for ECG Signal Analysis
CHAPTER 6 DISCUSSION ON RESULTS
CHAPTER 7 CONCLUSION AND FUTURE WORK
8 REFERENCES
9 LIST OF PUBLICATIONS
APPENDIX 1
APPENDIX 2
APPENDIX 3
APPENDIX 4
APPENDIX 5
JOURNAL PAPER
ACKNOWLEDGEMENTS
I thank the omniscient God for providing me his kindly light with love through the recesses and labyrinth of my research work.
The excruciatingly painful experience of writing a dissertation results in a unique reward. But this is not possible without the active support, help and cooperation of a good number of people. I therefore wish to express my deep sense of gratitude to each individual associated directly or indirectly in the successful completion of my research.
I am deeply indebted to my supervisor and mentor, Dr M. Madhavi Latha, Professor of ECE & Director, ITC, JNTUH, Hyderabad, for her immaculate guidance, untiring encouragement and inestimable inspiration. In fact, my devotion and reverence for her go beyond the pages of this volume.
It is a great pleasure to express my sincere thanks to Prof. Rameshwar Rao, Vice-Chancellor, JNTUH University, Hyderabad, Dr. A.Vinay Babu, Principal, JNTUH College of Engineering, Hyderabad for providing me an opportunity to fulfill my long cherished dream of pursuing research.
My sincere thanks are due to Dr. L. Pratap Reddy, Director, R & D Cell, JNTUH University, Hyderabad for providing me excellent guidance in the preparation of thesis. I thank Dr M. Madhavi Latha, Charmain BOS(ECE), JNTUH College of Engineering, Hyderabad for their valuable suggestions during the review period.
I wish to thank Prof P.S Raju, Director of GRIET, Hyderabad for his parental concern for me. I place on record my deep sense of appreciation for the support extended to me by Prof Jandhayala N Murthy, Principal of Gokaraju Rangaraju Institute of Engineering and Technology, Hyderabad.
A good number of the faculty of GRIET and JNTU have encouraged me in various ways during the course of my research. I wholeheartedly thank them for their help and support.
My thanks are due to my colleagues and students for their ready help and cooperation whenever required.
I appreciate the unstinted and ungrudging help, cooperation and understanding of my husband, Mr. T. Satish Kumar and my lovely tiny tots, Chi. Harshit and Chi. Druti. However, I find no words to express my love and gratitude to my Parents and Parents-in-law. How I wish them were alive today! (A.G. Padma)
ABSTRACT
Electrocardiogram records the electrical activity of heart. It is an effective diagnostic tool in identifying many cardiac abnormalities and it is being used for many years. According to the American Heart Association (AHA), cardiovascular disease is the main cause of mortality and morbidity in the world. There is a need for early detection of heart abnormalities to prevent further deterioration of the condition and to minimize unnecessary healthcare costs.
Early intervention during cardiac emergency can be life saving. Early detection of abnormalities can lead to early intervention and ECG plays a pivotal role in this aspect. The ECG can be used to diagnose evolving myocardial infarctions, identify potentially life-threatening arrhythmias, pinpoint chronic effects of sustained hypertension, and acute effects of massive pulmonary embolus. However, continuous monitoring is not possible with the traditional ECG machines. The electrode interface wires limit the portability of the instrument. During cardiac emergencies also the ECG is continuously monitored using the bed-side monitor, which restricts the movement of the patient.
The main purpose of the present work is to design and implement a prototype ECG system with wireless links for continuous monitoring of the subject for cardiac related problems. The ECG signal acquired from subject is filtered, digitized, and compressed for wireless communication. The ECG signal compression reduces the memory requirements for storage of the signal. This compression also decreases the burden on the communication system. The compressed data signal is transmitted through the wireless network (up to 100 meters). This replaces the cumbersome wiring system between sensing electrodes and the central monitoring station. Additional advantages of wireless communication include patient monitoring outside of hospital settling, thereby reducing the duration of hospitalization. The proposed system can be extended, upon interfacing with other devices, for continuous monitoring of other vital parameters of the patient.
In automation of the ECG signal analysis, the workload of the medical professionals can be reduced. The automated system provides an alert when critical changes are detected by the system. Concisely stated, ECG of the patient is continuously monitored and deviations from normalcy are detected in real-time. The changes in the ECG could be due to heart attack, fibrillation or arrhythmias. In case of emergency, data is transmitted to a medical practitioner, who in turn can provide necessary directions to take care of the situation. In this manner, as the problems can be detected as and when they occur, the remedial actions are initiated before the problems become serious.
The complete ECG diagnostic system includes a low power Instrumentation amplifier, filters, ADC, Microcontroller and ZIGBEE modules. MATLAB / LABVIEW are used for signal analysis and classification. These environments are capable of not only collecting, recording, transmitting, and displaying ECG data on a real time basis but also for analyzing the acquired ECG data in order to detect the cardiac abnormalities.
The MIT-BIH database signals were used for validation and evaluation of classification algorithms. In order to reduce the memory requirements for storing the acquired ECG signals, ECG data was compressed. Discrete Cosine Transform (DCT) technique was applied for ECG data compression. Here DCT showed good performance with a Compression Ratio (CR) of 82-90.43% and Percent Root Mean Difference (PRD) of 7.9-0.93. Linear Vector Quantization method (LVQ) is used for identifying the abnormalities associated with the ECG signal. After training the LVQ process with a reasonable number of samples, the algorithm is used for classifying ECG signals. The techniques used in the present work for ECG signal compression and classification gave better results compared to those found in the literature.
LIST OF TABLES
Table 1.1 Amplitudes and Duration of a Normal ECG Signal
Table 2.1 Heart Rate of subjects (only sample data)
Table 3.1 Comparison of Various Compression techniques
Table 3.2 Comparison of DCT Compression by different Authors
Table 4.1 Number of cases for various category
Table 4.2a LVQ weights relation
Table 4.2b LVQ Classification relation
Table 4.2c Initial Weight vector
Table 4.2d Weight vector after 100 epochs
Table 4.3 Performance classification system based on LVQ of the test phase
Table 5.1 Comparison of results with various thresholds, CR, PRD and Error
Table 5.2 DCT Compression applied on Supra Ventricular Arrhythmia
Table 5.3 DCT Compression applied on Ventricular Tachycardia 123 Table 5.4 DCT Compression applied on Malignant Ventricular Arrhythmia
Table 5.5 DCT Compression applied on Normal ECG
Table 5.6 Summary of Tables 5.2, 5.3, 5.4 and 5.5
Table 5.7(a) Data Classification for Class 1 and Class 2
Table 5.7(b) Data Classification for Class 3 and Class 4
Table 5.8 Comparison of the various ECG analysis methods from different Authors
LIST OF FIGURES
Fig. 1.1 ECG Generation
Fig. 1.2 Normal ECG
Fig. 1.3 ECG Lead System
Fig. 1.4 Basic Block diagram of Proposed ECG system
Fig. 2.1 Information Flow Diagram of ECG System implementation
Fig. 2.2 Metal plate electrodes
Fig. 2.3 AD620 Instrument Amplifier Pin Diagram
Fig. 2.4 ECG data acquisition system circuit
Fig. 2.5 Second order Low Pass Bessel Filter circuit
Fig. 2.6 High Pass Filter circuit
Fig. 2.7 Twin T network for Notch Filter
Fig. 2.8 Power supply Circuit
Fig. 2.9 Transmitter and receiver circuit of ZigBee
Fig. 2.10 Digital Data in Hyper terminal
Fig. 2.11 Delta Modulation
Fig. 2.12 First order Sigma delta ADC
Fig. 2.13 Second Order Sigma Delta ADC
Fig. 2.14 Second order Sigma delta modulator using Simulink
Fig. 2.15 Model based Design Simulation Result, Hysterisis Curve and Frequency response of Sigma Delta 2 nd order ADC
Fig. 2.16 Block Diagram of Proposed ECG Analysis System
Fig. 2.17 Data Acquisition Module
Fig. 2.18 Amplification Module
Fig. 2.19(a) Filtering part of ECG Analysis system
Fig. 2.19(b) Output Waveform at Filtering Module
Fig. 2.20 (a) QRS Detection and Heart Rate Calculation Module
Fig. 2.20 (b) Output waveform of QRS Detection & Heart Rate Calculation
Fig. 3.1 (a) Flowchart of AZTEC Compression for Line Detection Method
Fig. 3.1 (b) Flowchart of AZTEC Compression for Line Processing Method
Fig. 3.2 AZTEC Compression Applied on ECG Signal
Fig. 3.3 Flowchart of Turning Point Algorithm
Fig. 3.4 Results of Turning Point Compression Method
Fig. 3.5 Flowchart for CORTES Algorithm
Fig. 3.6 CORTES Compression Applied on ECG Signal
Fig. 3.7 Flowchart for DCT Compression Algorithm
Fig. 3.8. DCT compression Applied on ECG Signal
Fig. 3.9 Flowchart for FFT Compression Algorithm
Fig. 3.10 FFT Compression of ECG Signal
Fig. 4.1 Derivative Output of y(nT) ECG Signal
Fig. 4.2 Squaring Output of ECG Signal
Fig. 4.3 QRS Width using Window Integration
Fig. 4.4 R-R interval
Fig. 4.5 ECG waveform showing ST segment
Fig. 4.6 Functional description of a single Neuron
Fig. 4.7 A 3-2-2 configured multilayer perceptron
Fig. 4.8 LVQ symbolic structure
Fig. 5.1 ECG data acquisition setup and outputs at various stages of ECG amplifier
Fig. 5.2 Abnormal and Normal ECG waveforms
Fig. 5.2 Output Results for Bradycardia signals using LVQ Method
Fig. 5.3 Output Results for Myocardial Infarction using LVQ Method
Fig. 5.4 Output Results for Tachycardia using LVQ Method
NOTATIONS
Abbildung in dieser Leseprobe nicht enthalten
CHAPTER 1 INTRODUCTION
India has the highest incidence of heart related diseases in the world and the number of people affected is likely to double in coming years, according to the Indian Medical Association studies. The chest pain, shortness of breath, palpitation and hypertension are the symptoms of Cardio Vascular Disease (CVD). These are due to disorders of the heart and blood vessels which may even lead to death. More than 1.4 million people are suffering from angina and out of them 2, 75,000 people annually had heart attack. Majority of health institutes suggested that people with arrhythmias in emergency as well as in elective settings should be given timely review by an appropriate clinician to make sure of exact diagnosis, effectual treatment and rehabilitation.
1.1 ECG AND ITS IMPORTANCE
The heart is a muscular organ responsible for pumping blood through the blood vessels by repeated and rhythmic contractions in human beings. The average human heart, beating at 72 beats per minute, will beat approximately 2.5 billion times during a lifetime (about 66 years). It weighs on average 250g to 300g in females and 300g to 350 g in males.
The function of the right side of the heart is to collect deoxygenated blood, in the right atrium, from the body (via superior and inferior vena cava) and pump via the right ventricle, into the lungs (pulmonary circulation) through pulmonary valve so that carbon dioxide is exchanged with oxygen. This happens through the passive process of diffusion. The left side collects oxygenated blood from the lungs into the left atrium. From the left atrium the blood moves to the left ventricle which pumps it out to the body (via the aorta). On both sides, the ventricles are thicker and stronger than the atria 2.
The Sino Atrial (SA) node is the natural pacemaker that regulates the cardiac function. The SA node is located at the upper portion of the Right Atrium (RA) and is a collection of specialized electrical cells. SA node generates the pulses at regular intervals that travel through a specialized electrical pathway and stimulates the muscle wall of the four chambers of the heart to contract in a certain sequence or pattern. The upper chambers or atria are first stimulated. This is followed by a slight delay to allow the two atria to empty. Finally, the two ventricles are electrically stimulated to expel the blood into the arteries.
As the SA node fires, each electrical impulse travels through the right and left atria. This electrical activity causes the two upper chambers of the heart to contract. This electrical activity can be recorded from the surface of the body as a "P wave" on the Electro Cardio Gram recording (ECG). The electrical impulse then moves to an area known as the Atria-Ventricular (AV) node. This node is located just above the ventricles. Here, the electrical impulse is held up for a brief period. This delay allows the right and left atrium to continue emptying the blood into the respective ventricles. This delay is recorded as "PR interval." The AV node thus acts as a "relay station" delaying stimulation of the ventricles long enough to allow the two atria to finish emptying. Following the delay, the electrical impulse travels through both ventricles (via special electrical pathways known as the right and left bundle branches). The electrically stimulated ventricles contract and blood is pumped into the pulmonary artery and aorta. This electrical activity is recorded from the surface of the body as a "QRS complex". The ventricle then recovers from this electrical stimulation and generates an "ST segment" and T wave on the ECG as shown in Fig 1.1.
Abbildung in dieser Leseprobe nicht enthalten
Fig 1.1. ECG Generation
Electrical impulses in the heart originate in the Sino-Atrial (SA) node and travel through the heart muscle where they impart electrical initiation of systole or contraction of the heart. The electrical waves can be recorded by electrodes placed on the skin. Electrodes are placed on different locations of the body to obtain the activity from different parts of the heart muscle. An ECG displays the voltage between pairs of these electrodes, and the muscle activity that they measure, from different directions, also known as vectors. This display indicates the overall rhythm of the heart and weaknesses in different parts of the heart muscle. The electrocardiogram is composed of waves and complexes as shown in the Fig 1.2. Waves and complexes in the normal sinus rhythm are the P wave, PR Interval, PR Segment, QRS complex, ST Segment, QT Interval and T Wave.
The P wave corresponds to both the atrium contracting (depolarizing) and priming the ventricle with blood. The QRS complex is where the stronger ventricles fire, pushing blood through the pulmonary artery to the lungs from right ventricle and through the aortic valve from left ventricle to the entire body. The T wave occurs during the time at which the ventricles repolarize themselves for the next beat. Normally, frequency range of ECG signal is of 0.05 - 100 Hz and its dynamic amplitude range is of 1 - 10 mV.
A typical ECG tracing of a normal heartbeat (or cardiac cycle) consists of a P wave, a QRS complex and a T wave. A small U wave is normally visible in 50 to 75% of ECGs. The baseline voltage of the electrocardiogram is known as the isoelectric line. Typically the isoelectric lines are measured as the portion of the trace following the T wave and preceding the next P wave.
Any ECG signal provides two kinds of information. The duration of the electrical wave crossing the heart which in turn decides whether the electrical activity is normal or slow or irregular and the second is the amount of electrical activity passing through the heart muscle which enables us to find whether the heart muscle is working in a synchronized manner or not.
Abbildung in dieser Leseprobe nicht enthalten
Fig 1.2. Normal ECG
The normal value of heart beat lies in the range of 60-80 beats/minute. A slower rate is called Bradycardia (Slow heart rate) whereas higher rate is called Tachycardia (Fast heart rate). If the cycles are not evenly spaced, an arrhythmia may be indicated and a PR interval greater than 0.2 seconds can cause blockage of the AV node.
Table 1.1 provides the amplitude and duration of different waves of the ECG waveform3. These values are normal values and if deviated from these standard values it indicates a physiological problem.
Table 1.1. Amplitudes and Durations of a Normal ECG Signal.
Abbildung in dieser Leseprobe nicht enthalten
1.2. ECG LEAD SYSTEM
An electrocardiogram is obtained by measuring electrical potential between various electrodes placed on the body using an Instrumentation Amplifier of the Biomedical grade. The electrical signals of the heart from a particular combination of recording electrodes, which are placed at specific points on the surface of human body, are referred to as leads as shown in Fig 1.3.
Abbildung in dieser Leseprobe nicht enthalten
Fig 1.3 ECG Lead system
There are two types of leads systems: unipolar lead system and bipolar lead system. The unipolar lead systems have a indifferent (reference) electrode at the center (neutral) of the Einthoven's triangle which is located at zero potential. The direction of these leads is from the “center” of the heart radially outward. These include the precordial (chest) leads and augmented limb leads: AVR (Augmented Vector Right), AVL (Augmented Vector Left) and AVF (Augmented Vector Foot). In bipolar type both electrodes are at some potential, with the direction of the corresponding lead being from the electrode at lower potential to the one at higher potential, e.g., for limb lead I, the direction is from left to right. The limb leads are: Lead I, Lead II, and Lead III.
The 12 lead ECG system consists of the following:
1. Lead I is a dipole with the negative electrode on the right arm and the positive electrode on the left arm.
2. Lead II is a dipole with the negative electrode on the right arm and the positive electrode on the left leg.
3. Lead III is a dipole with the negative electrode on the left arm and the positive electrode on the left leg.
4. Lead AVR or "Augmented Vector Right" has the positive electrode on the right arm. The negative electrode is a combination of the left arm electrode and the left leg electrode, which "augments" the signal strength of the positive electrode on the right arm.
5. Lead AVL or "Augmented Vector Left" has the positive electrode on the left arm. The negative electrode is a combination of the right arm electrode and the left leg electrode, which "augments" the signal strength of the positive electrode on the left arm.
6. Lead AVF or "Augmented Vector Foot" has the positive electrode on the left leg. The negative electrode is a combination of the right arm electrode and the left arm electrode, which "augments" the signal of the positive electrode on the left leg.
7. Lead V1 is placed in the fourth intercostal space to the right of the sternum.
8. Lead V2 is placed in the fourth intercostal space to the left of the sternum.
9. Lead V3 is placed directly between leads V2 and V4.
10. Lead V4 is placed in the fifth intercostal space in the mid clavicular line (even if the apex beat is displaced).
11. Lead V5 is placed horizontally with V4 in the anterior axillary line.
12. Lead V6 is placed horizontally with V4 and V5 in the mid-axillary line.
The ground (reference) electrode is placed on the right leg. The ECG signal is extremely valuable for the cardiologist to evaluate the heart conditions like irregular heartbeats, occurrences of heart attack, and damage to the parts of heart.
1.3 NEED FOR ECG MONITORING SYSTEM
The Electrocardiogram (ECG) is the most popularly adopted clinical tool that can evaluate and record the electrical activity of the heart. During the diagnostic ECG recording there is a chance to miss rare or intermittent symptoms, as these symptoms may not appear during the brief session of recording. To overcome this, the present existing solution is Holter monitoring system. During Holter monitoring, continuous recording of ECG signal is performed, as the patient is performing his/her normal routine activities except bathing. It is used to record an individual's ECG data for 24-48 hours, though the useful or critical information with Holter monitoring are very low (5-13%) 4, as the episode time is of the order of few seconds to few minutes. The episode time is very small compared to the entire recording, and this makes the system inefficient as far as memory requirement is concerned. After a wide range of survey for a variety of commercially available monitoring systems as well as some of the most visible research proposals, the existing systems can be classified into three groups2: a) systems that record signals and perform classification off-line, b) systems that perform remote real-time classification and c) systems that provide local online real-time classification.
The improvement of Holter-based monitors and event recorders stand out in the group (a) of systems like GE's SEER, Philips's DigiTrack, and Midmark's IQmark. The basic drawback in these solutions of recording devices is: there is no real-time processing of ECG and analysis, if any, is mostly performed off-line. The group (b) augments the telemetry related functionalities via remote real-time monitoring 5678911. In this case most of them make use of Mobile Phones or Personal Digital Assistance (PDA) to collect the ECG data and transmit it to a central monitoring station, where the processing of ECG can take place on a PC and classification of arrhythmias, if any, is performed by a clinical expert. The group (c) includes recent research proposals that can provide some intermediate level of local real-time ECG processing and/or classification of arrhythmias by using up-to-date smart phones or Personal Digital Assistance 121314. The initial results of these efforts show high promise to reduce the turnaround time. However several technological challenges must be addressed before a truly mobile cardiovascular disease diagnostic solution can be realized.
1.4 LITERATURE SURVEY
The implementation of a heart monitor using Analog Devices 16 involves a low cost, low power instrumentation amplifier and filter components coupled with a sophisticated microcontroller and LCD screen. This device is portable and consumes very less power. Since the ECG signal is a low amplitude signal, it is to be amplified against noise emanating from other muscles and electrical sources. The display of the heart rate is obtained by measuring the time between two successive R waves, and then calculating the heart rate i.e. Beats Per Minute (BPM).
The high resolution ADC with wide signal bandwidth has become an essential component for advanced biomedical analysis. Because the frequency of physiological signal is low compared with other general signals, the Sigma Delta Modulator (SDM) is the most appropriate ADC architecture for Biomedical Applications.
The modular nature of virtual instrumentation can be easily added to a new functionality. LABVIEW implementation uses dataflow programming and the flow of data determines the execution as discussed by Jun Hao Yu 18 in ECG analyzing experimental system based on virtual instrument.
Vergari 19 presented a ZigBee-based ECG transmission system for home care. The wireless data transmission using ZigBee has the advantages of low power transmission, high transmission range (100feet), and more number of nodes can be interconnected to increase the distance of transmission range. It is a low cost solution and can be used for providing high quality services to older people like e-services at home.
Belgachem 20 discussed a supervised classification of ECG using LVQ Neural Network with 32 elements of information for classification, out of which 30 elements are used for QRS segments and next two for instantaneous RR and QRS complex, which are used to identify only two most common ECG waveforms like normal and PVC (Premature Ventricular Contraction) from the data obtained from MIT/ BIH database.
According to Anuradha 21 in Cardiac Arrhythmia Classification using Fuzzy Classifiers the accuracy obtained was 93.13%. The four parameters considered for cardiac arrhythmia classification using ANN are Spectral entropy, Poincaré plot geometry, Largest Lyapunov exponent and Detrended fluctuation analysis. A total of 228 rules were used in this work for eight cardiac arrhythmia classifications. The difficulty of real time implementation, in case of a large number of fuzzy rules, was also discussed.
Kabdi 22 has implemented wavelet transform and artificial neural network based classification, in which three kinds of features are computed by Joint time-frequency feature (Discrete Wavelet Transform Coefficients), Time domain feature (R-R intervals) and Statistical feature (form factor) and achieved an overall accuracy of 90%.
Kannathal 23 used three non-linear parameters as inputs to an adaptive neuro-fuzzy network classifier for classification of different types of arrhythmias with an accuracy level of 94%. Rodrigo 24 has implemented the perceptual masking method to compress ECG signals by threshold, numerical masks to Discrete Cosine Transform (DCT) coefficients and achieved average Compression Ratio(CR) of 52.42% and average Percent Root mean Difference(PRD) of 1.24.
Bouali 25 applied an orthogonal transform on a window of ECG signal, and then proceeded to reduce the number of bits representing transform coefficients under threshold of a DCT compression algorithm based on optimal bits allocation. The CR of 3% and PRD of 5 was achieved.
Leonardo 26 implemented ECG compression based on optimized quantization of Discrete Cosine Transform (DCT) coefficients. The ECG to be compressed is partitioned in blocks of fixed size, and each DCT block is quantized using a quantization vector and a threshold vector that are specifically defined for each signal. Then the vectors are defined, via Lagrange multipliers and estimated entropy is minimized for a given distortion in the reconstructed signal. The ECG signals are adopted by the quantized coefficients which are coded by an arithmetic coder. By the above implementation the average CR of 9.3% and PRD of 2.5 was achieved.
Ahmed 27 implemented a new hybrid two-stage ECG signal compression method based on the Modified Discrete Cosine Transform (MDCT) and Discrete Wavelet Transform (DWT), in which the ECG signal is partitioned into blocks and the MDCT is applied to each block to de-correlate the spectral information. Then, the DWT is applied to the resulting MDCT coefficients. Removal of spectral redundancy is achieved by compressing the subordinate components more than the dominant components. Resulting wavelet coefficients were thresholded and compressed using energy packing and binary significant map coding technique to obtain CR of 21.5% and PRD of 5.89.
1.5 METHODOLOGY
The present work seeks to construct a wearable system, capable of performing continuous monitoring and recording of ECG in real time, by automatically detecting arrhythmias and classifying them in real time at any place and any given time. After an extensive literature review it is found that there is no report of a complete ECG monitoring capable of rhythm classification using advanced machine learning algorithms. Moreover, most of prior portable solutions proposed rule-based CVD (Cardio Vascular Disease) detection or simple clustering techniques, which lack the individual adaptability to provide more accurate assessment. Thus the focus is on making the following contributions in this thesis work: a) Construction of a portable physical prototype that includes an off-the-shelf wireless ZigBee module for heart monitor. The MATLAB/ LABVIEW simulation environment is capable of not only collecting, recording, transmitting, and displaying ECG data in real time but also analyzing the acquired ECG data and detecting cardiac abnormities. The proposed system is shown in Fig 1.4. b) Design of machine learning based CVD diagnosis module and implementing an adaptive Artificial Neural Network (ANN) based patient-specific training methods and establishing medical database.
MATLAB / SIMULINK / LABVIEW
Abbildung in dieser Leseprobe nicht enthalten
Fig 1.4. Basic Block diagram of proposed ECG system
The following Hardware modules and algorithms are implemented in the present work:
- Instrumentation Amplifier (AD624), OP Amps (LM 741), ADC (ADC0804), Microcontroller (ATMEL 89C51) and ZigBee based wireless transmission.
- Design of a second order Sigma Delta ADC.
- QRS detection and Heart Rate Monitoring.
- AZTEC, TP, CORTES, FFT and DCT compression techniques.
- Analysis of ECG signals by LVQ network.
The main purpose of the work is to design the ECG amplifier with associated filter circuits, signal compression using the Discrete Cosine Transform and classification using Linear Vector Quantization.
Though several methods were available in the literature, the techniques presented in this present work gave better performance. The workings of different parts of the system constructed are explained below.
An instrument is designed for acquiring the real time ECG data from subjects. The preamplifier is a low cost and low power, monolithic instrumentation amplifier (AD624), manufactured by Analog Devices. The gain of the instrumentation amplifier is set with an external resistor. The gain can be set in the range of 1 to 10,000. The instrumentation amplifier has high common mode rejection ratio (CMRR), high source resistance of 1M ohm, low nonlinearity and withstands input overloads of upto + 15v or 60mA for several hours.
The low pass, high pass and notch filters are designed using opAmp LM741. The design details are provided in subsequent chapters.
The ECG signal thus acquired is input to a microcontroller based system, with ADC interface. The AT89C51 microcontroller having 128 bytes of RAM, 4 K bytes of on-chip ROM, two timers, one serial port and four I/O ports (each 8 bits wide), is used in the present work. As the microcontroller does not have a built-in ADC, an external ADC is used. The National semiconductor produced ADC 0804 is used. The ADC0804 IC is an 8 bit successive approximation register type ADC with conversion time of 110 gs. This ADC is microcontroller compatible and can be directly interfaced to the microcontroller.
ZIGBEE is a low-power wireless sensor network for transmission. This operates in the frequency of 2.4 GHz for indoor application up to 100ft (30 m), Outdoor line-of-sight up to 300ft (100 m). It has Transmission Power of 1 mW with a data transmission rate of 250Kbps.
ADC, in addition to hardware construction, is also implemented using SIMULINK for Sigma Delta ADC. Advantage of Sigma-Delta ADCs is that they do not require high-precision and accurately trimmed analog components for high resolution and high integration. Sigma-Delta Modulator (or noise-shaper) shapes the quantization noise and pushes the majority of the in band noise to higher frequencies. The digital filter removes the high frequency quantization noise and down samples the high frequency to Nyquist rate. The techniques of over-sampling and noise shaping allow the use of relatively imprecise analog circuits to perform high resolution conversion using only 1-bit A/D converter.
LABVIEW is a graphical programming language used for creating test, measurement and automation. It provides flexibility in developing an ECG analysis system based on virtual instrumentation. In the present work LABVIEW was used for experimental investigation to guide the implementation of hardware module.
Compression Ratio (CR) and Percent Root Mean Difference of Error (PRD) are used as criteria to evaluate the performance of the applied compression technique. The Discrete Cosine Transform resulted in good compression for all the coefficients above the threshold of 0.19. Average Compression Ratio of 82-90.43% and PRD of 0.93-7.9 were obtained with coefficients above threshold value. Additional tests, using the same threshold, on various arrhythmia signals like Supra Ventricular Arrhythmia, Ventricular Tachycardia, Malignant Ventricular-Arrhythmia (Ventricular Tachycardia, Ventricular Flutter, and Ventricular Fibrillation) using MIT-BIH database resulted in an average CR of 85% and PRD of 6.2.
LVQ method is used for identifying the four classes of arrhythmias: Tachycardia, Bradycardia, Pre-mature Ventricular Contraction and Myocardial Infarction. QRS complex duration, R-R interval, R-wave amplitude, ST segment slope change are the four features used as weights in LVQ calculations. Vector quantization divides the input space into areas that are assigned as “code book” vectors based on supervised learning. During the training process, the output units are positioned to approximate the decision surfaces. After training, an LVQ net classifies an input vector by assigning it to the same class as the output unit that has its weight vector (reference or codebook vector) closest to the input vector based on Euclidean distance.
The LVQ network, when trained with 140 samples with a learning rate of 0.9, gave the best performance with an overall accuracy of 95.5%.
1.6 ORGANIZATION OF THESIS
The thesis is organized into eight chapters as follows:
The first chapter deals with the Introduction of ECG system, its importance and historical study of existing ECG systems, which lead to the discovery of certain lacunae in the current monitoring systems.
The second chapter deals with the hardware implementation of ECG system, the calculation of heart rate using LABVIEW software and the transmission of ECG signal using ZIGBEE.
The third chapter deals with ECG compression, its implementation, comparisons of various techniques and identification of their relative advantages.
The fourth chapter deals with the ECG analysis using LVQ technique and its implementation, detection of arrhythmias and its classifications.
The fifth chapter deals with the experimental results of the hardware for data acquisition, compression and analysis.
The sixth chapter summarizes all the above chapters and presents the main conclusions,
The final chapter summarizes all the above six chapters and offers certain recommendations for further research in this vital area of cardiology.
CHAPTER 2 ECG DATA ACQUISITION SYSTEM AND HEART RATE MEASUREMENT
The hardware of the proposed system consists of different modules for signal acquisition, transmission and analysis. The heart rate calculation is performed with the help of LABVIEW software. The detailed discussion on the implementation is as follows:
2.1 ECG SYSTEM REQUIREMENTS
The proposed heart rate measurement system consists of four functional units. These functional units perform the signal acquisition, processing, communication and analysis. The functional units are:
a. Data Acquisition Unit.
b. Data Processing Unit.
c. Data Communication Unit.
d. Data Analysis Unit.
These four functional units are briefly described in the following sections:
2.1.1 Data Acquisition Unit
The data acquisition unit consists of electrodes placed on the surface of the body, which can continuously pickup the ECG signal and provide an output in the form of an electrical signal. The acquired signal, owing to its low amplitude, needs to be amplified using an instrumentation amplifier. The electrodes, also pick-up noise along with the required ECG signal. The noise signals include Baseline drift/wander, power-line interference and high frequency noise from high frequency signal generators in the vicinity of the subject. Thus, it is necessary to filter the preamplifier output using low pass filter, high pass filter and notch filter. The processed ECG signal is then fed to the data processing unit.
2.1.2 Data Processing Unit
The acquired and filtered ECG signal is applied to the data processing unit. The data processing unit consists of a Microcontroller with sufficient memory. The analog ECG signal is fed to an Analog-to-Digital Converter, which is interfaced to the microcontroller. The ECG signal is continuously sampled at regular intervals and the digitized values are stored in the RAM area associated with the microcontroller. The microcontroller transmits, these samples to PC, where a MATLAB program receives these values and the ECG signal samples are plotted on the PC for viewing and also for subsequent analysis.
2.1.3 Data Communication Unit
The data communication unit is essentially based on ZIGBEE technology, since it works at a Radio Frequency (RF) specification for short-range, point-to-point and point-to-multipoint for voice and data transfer applications. ZIGBEE will enable users to connect to a wide range of computing and telecommunication devices. The acquired ECG data is transmitted to a nearby physician for further analysis and for identification of the associated abnormalities. This real time communication allows for immediate remedial action that may be necessary.
Thus, this whole device can be deployed as a wearable device, just by modifying the electrode system. In order to avoid moving artifacts and to provide proper contact with the body wearable electrodes can be used, by modifying the preamplifier suitably.
2.1.4 Data Analysis Unit
The data analysis unit is a MATLAB based software package for ECG signal analysis. This unit deals with analysis of ECG signal. The concept of this analysis is based on QRS detection and its analysis. The software program developed is capable of identifying QRS complexes from the raw ECG data. The output of this unit provides various ECG parameters for further investigations. The information flow diagram of ECG system implementation is shown in Fig 2.1.
Abbildung in dieser Leseprobe nicht enthalten
Fig 2.1 Information Flow Diagram of ECG System Implementation.
2.2 ECG SIGNAL DATA ACQUISITION
Electrocardiograph (ECG) is a trans-thoracic interpretation of the electrical activity of the heart over time, captured and externally recorded by using skin (surface) electrodes. It is a non-invasive recording produced by an electrocardiographic device. The ECG identified mostly by detecting and amplifying the tiny electrical changes on the skin that are caused when the heart muscle “depolarizes” during each heart beat. At rest, heart muscle has a charge across its outer wall or cell membrane. Reducing this charge towards zero is called depolarization, which activates the mechanism in the cell that causes it to contract, i.e. producing mechanical action. During each heart beat a healthy heart will have an orderly progression of a wave of depolarization that is triggered by specialized cells located in the Sinoatrial (SA) node, spreads out through the atrium, passes through “intrinsic conduction pathways” and then spread all over the ventricles. This is detected as tiny rises and falls in the voltage between two electrodes placed suitably on the body which is displayed as a wavy line either on a screen or on a paper. This display indicates the overall rhythm of the heart and associated mechanical activity in different parts of the heart muscle 10. The deviation in this waveform can be used to identify the problems associated with the heart muscle.
Usually, more than two electrodes are used and they can be combined into a number of pairs (for example: left arm (LA), right arm (RA) and left leg (LL) electrodes form the three pairs LA-RA, LA-LL and RA-LL). The output from each pair of electrodes is known as a lead. Each lead is said to look at the heart from a different angle.
It is the best way to measure and diagnose physiological abnormalities of heart particularly abnormal rhythms caused by damage to the conductive tissue that carries electrical signals, or abnormal rhythms caused by electrolyte imbalances.
2.2.1 Electrodes used for ECG Signal Pickup
The mechanism of electrical conductivity in the body involves ions as charge carriers. Thus, picking up bioelectric signals involves interacting with these ionic charge carriers and transducing ionic currents into electric currents required by electronic instrumentation. This transducing function is carried out by electrode that consists of electrical conductors in contact with the aqueous ionic solution of the body. The interaction between electrons in the electrodes and ions in the body can greatly affect the performance of these sensors and requires careful considerations.
The electrodes are placed on the skin surface, using conductive gel or paste. The purpose of conductive gel or paste is to provide better contact. ECG electrode contact impedance on dry skin is 100 Kilo Ohms and the equivalent capacitance is 0.01 micro farads. But, with the application of electrode gel, the contact impedance is 10 Kilo Ohms and capacitance is 0.1 microfarads. Thus, use of electrode gel reduces the electrode contact resistance thereby reducing the moving artifacts. Many different forms of electrodes have been developed for different types of electro-physiological measurements. The relevant electrode types useful for ECG recording are considered and described in the following sections.
2.2.1.1 Surface Electrodes
The surface electrodes can be placed on the body surface for recording bioelectric signals. The integrity of the skin is not compromised when these electrodes are applied and they can be used for short term diagnostic recording such as taking a clinical electrocardiogram or long term chronic recording such as in cardiac monitoring.
The basic metal plate electrode consists of a metallic conductor in contact with the skin with a thin layer of an electrolyte gel between the metal and the skin to establish better contact. Normal ECG recording metal plate electrodes are shown in Fig. 2.2a and b.
Abbildung in dieser Leseprobe nicht enthalten
a) Limb electrodes b) Cup electrodes
Fig. 2.2 Metal plate electrodes
Metals commonly used for this type of electrodes include German-silver (nickel-silver alloy), silver-silver chloride, Gold and Platinum. Sometimes these electrodes are made of metal foil so as to be flexible and sometimes they are produced in the form of suction electrode, to make it easier to place electrodes on the chest.
2.2.1.2 Adhesive Electrode
The pressure of the surface electrode against the skin may squeeze the electrode paste out. To avoid this problem adhesive tape is used. It consists of a lightweight metallic screen backed by a pad for electrode to attach. The adhesive backing holds the electrode in place and retards the evaporation of the electrolyte present in the electrode paste. The disposable adhesive electrodes are attached to the patient's skin and can be easily removed.
2.3 INSTRUMENTATION AMPLIFIER
The electrocardiograph signal needs to be conditioned before being processed through the complete circuit. The Instrumentation amplifier is used to amplify the ECG signal picked up by the electrodes.
2.3.1 Requirements of Instrumentation Amplifier
The biomedical instrumentation amplifier requires the following specifications:
a. The voltage gain of the amplifier should be more than 100dB so as to amplify the bio-signal adequately.
b. The gain should be uniform throughout the required bandwidth.
c. The input impedance should be very high.
d. The output impedance should be very small.
e. The CMRR should be more than 80dB so as to eliminate the noise.
In order to meet all these specifications, the preamplifier i.e. instrumentation amplifier is used. Instrumentation amplifier can be constructed using discrete components or one can even use an instrumentation amplifier. The first stage in both the cases is a differential amplifier.
2.3.2 AD620 Instrumentation Amplifier
The AD620 as shown in Fig 2.3 is a low cost, high accuracy instrumentation amplifier that requires only one external resistor to set gain in the range of 1 to 10,000. It offers low power dissipation (with only 1.3 mA max current), making it a better choice for battery- powered and portable applications.
The AD620, with its high accuracy of 40 ppm maximum nonlinearity, low offset voltage of 50 pV max, and offset drift of 0.6 pV/°C max, is ideal for use in precision data acquisition systems like transducer interfaces. The low noise, low input bias current (1.0 nA max), and low power of the AD620 make it well suited for medical applications [Appendix 1].
Abbildung in dieser Leseprobe nicht enthalten
Fig 2.3 AD 620 Instrumentation Amplifier pin Diagram.
The AD620 works well as a preamplifier due to its low input voltage noise of 9 nV per each 1Hz at 1 KHz, 0.28 pV p-p in the 0.1 Hz to 10 Hz band, 0.1 pA for each 1Hz input current noise with its settling time of 15 ps to 0.01%. This amplifier has an excellent DC and AC performance. The internal gain resistors, R1 and R2, are trimmed to an absolute value of 24.7 KQ, allowing the gain to be programmed accurately with a single external resistor.
The gain equation is given as:
Abbildung in dieser Leseprobe nicht enthalten
Where Rg - Gain choosing Resistor, G - Required gain.
The required gain is first selected, and based on the required gain, the gain choosing resistor value is calculated using the mathematical equation (2.2). In the present application the gain of this amplifier is chosen as 1000. With very high gain, the amplifier may get saturated with higher amplitude noise signals. If the gain is too low, the signal itself is not amplified properly. Keeping this in view, the amplifier gain is selected as 1000.
2.4. SIMPLE ECG ACQUISITION SYSTEM
The low current noise of the AD620 with high input resistances of 1 MQ, allows its use in ECG amplifiers. To avoid leakage current associated problems, a driven right leg configuration is used. Appropriate isolation circuits need to be used to provide proper safegauards and for further reduction in leakage currents. In the present circuit, a specific isolation circuit is not used. A good quality transformer with grounded shield is used to reduce the leakage current, and the same provides the required safety. The driven right leg configuration is provided by using AD705 as shown in the Fig 2.4. The main instrumentation amplifier AD620 features a series thin film resistance of 400Q at it input terminals, which makes the IC to withstand overloads of upto ±15V or ±60mA continously for several hours.
Abbildung in dieser Leseprobe nicht enthalten
Fig 2.4 ECG Data Acquisition System Circuit
2.4.1 Filters used in ECG System
The filtering section is designed as a part of preprocessing of ECG signal. This includes low pass, high pass and notch filters. Standard ECG components of interest will reside in the 0.67 to 100Hz bandwidth. The filter sections are to be followed immediately after the signal acquisition part to eliminate noise present in the ECG signal.
2.4.1.1. Low Pass Filter (LPF)
The desired high frequency limit of the ECG signal is 100 Hz.
The low pass filter restricts the noise above the 100 Hz from encroaching into the ECG system. Low pass filter was used to eliminate the unwanted high frequency noise, thereby limiting the ECG upper limit to 100 Hz. A second order Bessel filter is chosen due to the constant group delay with no overshoot. Designing a second order low pass Bessel filter with cutoff frequency of 100Hz and unity gain requires determination of resistor and capacitor values using mathematical equations (2.3) and (2.4). The calculated second order Bessel coefficients are: a= 1.732 and b=0.785, C 1 =220nF.
[...]
-
¡Carge sus propios textos! Gane dinero y un iPhone X. -
¡Carge sus propios textos! Gane dinero y un iPhone X. -
¡Carge sus propios textos! Gane dinero y un iPhone X. -
¡Carge sus propios textos! Gane dinero y un iPhone X. -
¡Carge sus propios textos! Gane dinero y un iPhone X. -
¡Carge sus propios textos! Gane dinero y un iPhone X. -
¡Carge sus propios textos! Gane dinero y un iPhone X. -
¡Carge sus propios textos! Gane dinero y un iPhone X. -
¡Carge sus propios textos! Gane dinero y un iPhone X. -
¡Carge sus propios textos! Gane dinero y un iPhone X. -
¡Carge sus propios textos! Gane dinero y un iPhone X. -
¡Carge sus propios textos! Gane dinero y un iPhone X. -
¡Carge sus propios textos! Gane dinero y un iPhone X. -
¡Carge sus propios textos! Gane dinero y un iPhone X. -
¡Carge sus propios textos! Gane dinero y un iPhone X. -
¡Carge sus propios textos! Gane dinero y un iPhone X. -
¡Carge sus propios textos! Gane dinero y un iPhone X. -
¡Carge sus propios textos! Gane dinero y un iPhone X. -
¡Carge sus propios textos! Gane dinero y un iPhone X. -
¡Carge sus propios textos! Gane dinero y un iPhone X. -
¡Carge sus propios textos! Gane dinero y un iPhone X. -
¡Carge sus propios textos! Gane dinero y un iPhone X. -
¡Carge sus propios textos! Gane dinero y un iPhone X. -
¡Carge sus propios textos! Gane dinero y un iPhone X. -
¡Carge sus propios textos! Gane dinero y un iPhone X. -
¡Carge sus propios textos! Gane dinero y un iPhone X. -
¡Carge sus propios textos! Gane dinero y un iPhone X. -
¡Carge sus propios textos! Gane dinero y un iPhone X. -
¡Carge sus propios textos! Gane dinero y un iPhone X. -
¡Carge sus propios textos! Gane dinero y un iPhone X. -
¡Carge sus propios textos! Gane dinero y un iPhone X. -
¡Carge sus propios textos! Gane dinero y un iPhone X. -
¡Carge sus propios textos! Gane dinero y un iPhone X. -
¡Carge sus propios textos! Gane dinero y un iPhone X. -
¡Carge sus propios textos! Gane dinero y un iPhone X.