ECG DATA COMPRESSION FOR TELEMEDICINE

Abstract

For critical cardiac patients, ambulatory patients, astronauts, person under cardiac surveillance and for creation of ECG database, the ECG signal is recorded continuously. Recorded data in these cases are so voluminous in size that it becomes practically impossible to handle it without compression. The data compression is also important as number ofcardiac patients is increasing at an alarming rate and it is not possible for the existing number ofcardiologists to take care ofall the cardiac patients spread throughout the world. Problem become more serious for the cardiac patients located in remote areas. Solution to the above mentioned problem is to transmit the ECG data of the patient to the cardiac expert centers through Internet. To handle the large volume of ECG data without losing the diagnostic information, it is necessary to use faithful data compression techniques. Thus, there is a need to develop such techniques, which have better performance in comparison to existing techniques. Although, large number of ECG data compression techniques have been developed in last four decade, but still no method can be claimed to reach a stage of perfection to deal with all type of ECG data of normal and abnormal subjects. There is ample scope of research to be carried out in this field, to make improvements in the existing data compression techniques and also to develop more efficient and effective new data compression techniques. Many methods and algorithms have been studied, defined and applied to perform data compression operations. As a general classification, they can be divided in two main groups: Information preserving or error-free (lossless) techniques and lossy techniques. The name itself is self-explanatory. In most of the ECG applications, the errorless methods do not provide sufficient compression, and hence, errors are expected in practical ECG compression systems. Because of the limited bandwidth, very low compression ratio will not be suitable for the telemedicine. Hence, lossy techniques for ECG data compression are preferred over lossless techniques. ECG data compression algorithms must also represent the data with acceptable fidelity. ECG data compression can also broadly classified into three categories i.e.: (i) Direct Data Compression (DDC) (ii) Transformation Compression (TC), and (iii) Parameter Extraction Compression (PEC). After dealing with the general introduction and the brief outline of the work, the first stage of the work deals with the direct data compression (DDC) methods as applied to ECG signals. These methods are generally preferred for ECG data compression over other XXI methods, because, they are easy to implement and require less computational time. The performance of these methods have been tested using CSE (Common Standards for quantitative Electrocardiography) and MIT/BIH (Massachusetts Institute of Technology/Beth Israel Hospital) databases. A more efficient method of ECG data compression has been developed using the basic concept ofexisting modified amplitude zone time epoch coding (MAZTEC) technique and named as improved modified AZTEC technique. With this method value of CR is 12.4 and PRD is 9%, whereas with modified AZTEC method CR is 10 and PRD is 14.2% for the same threshold value i.e. 0.035mV. The other DDC techniques namely, SAPA-1, SAPA-2, SAPA-3 (scan along polygonal approximation) and Fan along with residual coding technique have also been successfully implemented. In this new proposed method, it has been found that number of bits for transmission has reduced in all the cases. This reduction is of the order of 20%. Residual coding algorithm, which is based on the fact that the difference in amplitude between successive samples is typically smaller than the amplitude of samples, has been applied to all the above mentioned DDC techniques. Hence, the basic idea is that, when data samples are estimated, the error between the actual sample and estimated sample value is quantized and transmitted or stored. The residual coding algorithm shows better results as the number of bits required to transmit the compressed data are reduced (20% reduction in case of MA-001.DCD & MA-006.DCD for SAPA-1, 2, 3 and Fan at threshold 0.0lmV) with first difference coding. It has also been verified that the second difference coding does not improve compression as the number of bits remain same as to that of first difference signal. A low pass filter has been used to smooth the reconstructed signal. The effect of the length of the smoothing parabolic filter used during reconstruction has also been studied. A comparative study has been carried out which shows that seven point smoothing parabolic filter is the best to use for reconstruction of signal over nine and eleven point filters. It has also been observed that with the use of smoothing filters, high frequency noise such as power line interference and electomyographic noise are significantly reduced. A comprehensive study has been carried out on fidelity of the reconstructed signals. The performance has been evaluated on the basis of compression ratio, percent root mean square difference and the fidelity of the reconstructed signal. Further, in order to know the extend to which the diagnostic information is preserved during compression, features has been extracted using wavelet transform, both of the XXll original and reconstructed ECG signal. The comparison ofextracted parameters has been carried out to know the efficiency ofcompression. Insignificance of PRD as performance index in context ofdiagnostability has also discussed. Security ofthe ECG data has been incorporated by having ECG compressed data in authenticated header format. The problems related to bandwidth or channel capacity has been overcome by increasing the bit rate. It is expected that the implementation of these compression methods in real hospital environment would provide efficient solutions to ECG data transmission and storage. Next section ofwork deals with an efficient and composite algorithm, which has been developed for ECG data compression using error back propagation neural networks (EBP-NN). Four EBP-NN have been trained to retrieve all the 12 standard leads of the ECG signal. The combination of leads and the network topologies have been finalized after an extensive study ofcorrelation between the ECG leads using CSE database. Each network has a topology of N-4-4-N, where Nrepresents the number of samples in one cycle in any particular lead. The compression ratio in EBP-NN method goes on increasing with the increase in the number of ECG cycles (e.g. CR = 29.84 for 200 cycles, 470 samples in each cycles). This method is best suited for the data compression of nonarrhythmic ECG signals. The performance of the algorithm has been evaluated by comparing the vital reference points like onsets, offsets, amplitudes, durations and the extracted features of diagnostic significance like amplitudes, durations and ventricular activation time (VAT) of various wave segments of the original and the reconstructed ECG signals. The use of wavelet transform has been made for QRS detection, which sets the reference for the extraction ofreference points and the diagnostic parameters from the ECG signal. The test results at each stage are consistent and reliable. The test results and performance indices have proved beyond doubt that the EBP-NN method is very efficient for the data compression and help in the management of ECG data in both offline and real time applications. It has been observed that as long as the number of sample in one cycle is not varying too much, the EBP-NN method is working satisfactory; but as soon as the variation become significant, training of the network topology become more difficult and the quality of the retrieved signal is not good. Hence, it may result to misleading interpretation of the ECG signal. In other words, we can say that EBP-NN method is suitable for non-arrhythmia case where RR interval remains same. This may be considered XXIII as the limitation of the EBP-NN method of ECG data compression. Hence, a complete scheme for ECG data compression using adaptive EBP-NN and direct data compression (DDC) methods has been developed. Performance of the developed scheme has been checked on number of non-arrhythmic and arrhythmic cases and it worked successfully. A different approach has been used to check the performance of the retrieved signal i.e. section-wise error analysis using percent rms difference criterion. The quality of the retrieved signal is being evaluated before the transmission at the sending end itself, unless error is not fulfilling within limit criterion. This further reduces the chance of loosing any diagnostic information. In the second stage of the work, importance of ECG data compression in the area of telemedicine has been discussed. Telemedicine means the use of telecommunication and informatics in medicine. Telemedicine is a successful integration of medical expertise, medical equipment, computer hardware and software, telecommunication infrastructure and Internet into a system by which the patients can be examined, investigated, monitored and treated by medical expert from a distant place. Firstly, state of art of telemedicine has been presented. Because, the main work is centered around the ECG data compression hence, the telecardiology aspect has been dealt especially. This section gives details about current status and future prospect of telemedicine in context of the Indian health scenario. In general, the classification, mode of operation, advantages and challenges ahead to the telemedicine has been discussed. It is expected that, within the next few years no Indian will be deprived of a specialist consultation and will not be having any location disadvantage for specialized heath care service. The challenge today is not confined to overcoming technological barriers, insurmountable though they may appear, but as which technology is got to be implemented and at what cost. The bandwidth, latency, availability, security and ubiquity, which are essential requirements and posing challenges in effective implementation of telemedicine/telecardiology projects, have also been discussed. Continuous monitoring of the electrocardiogram and other signals related to current heart activity are necessary for patients who are suffering from cardiac diseases. Telecardiology provides a solution to the problem of healthcare for those cardiac patients who are in remote places and can not afford expanses and time to come to the places where experts are available. It is not only the patients who are being benefited, but at the same time the doctors also who can take care of a large number of patients without XXIV moving from their place. This work provides details ofatelecardiology system for use in a developing country. The backbone of the system is the Internet. Auser friendly web site has been designed, with the help a developed software. The compressed ECG data and other information related to the patient have been successfully uploaded. Aregistered doctor using software loaded on their client side downloads the patient data and information and decompresses it for further needful action. The importance of the data compression techniques has also been discussed. It has been shown that because of the limited bandwidth or channel capacity, it is essential to compress the ECG data before transmission. It has also been observed that the transmission time decreases with an increase in the compression ratio and also that, there is atrade-off between signal fidelity and compression ratio. The telecardiology scheme is working in store-forward mode of operation of telemedicine. A typical on-line feature on the designed website, is also provided for the doctors so that they may hold one-to-one discussion sessions with the patient or attending physician over a case underexamination. Finally, it can be stated that the work contributes significantly to the area ofECG data compression techniques. The developed methods are very much useful for telemedicine especially in telecardiology. It also raises number of questions for the carrying out further research work in this field. The overall work done in this thesis may be considered a positive and significant contribution for effective healthcare services to the remotely located patients.