ECG is a form of biomedical waveform that provid es a lot of necessary information to physicians. It reflects many aspects of physical conditions… [627120]

139
CHAPTER 5
CONCLUSION AND FUTURE SCOPE
5.1 Summary of Conclusions
ECG is a form of biomedical waveform that provid es a lot of necessary information to
physicians. It reflects many aspects of physical conditions which helps to record patients
biomedical data for a long period of time. To discover human body‘s abnormalities or
disorders, the 24 hour or even longer durat ion recording is desirable for doctors so as to
discover future clinical diagnosis or to transmit them for remote clinical diagnosis. The
recorded data is abundant in size, therefore ECG compression is necessary. For clinical
purposes, it is important to m aintain the diagnostic features of ECG signal during
compression.
In this thesis, transform based efficient quality controlled ECG compression techniques
are proposed and analyzed. The results are presented on different ECG signals of varying
characteristi cs.
First, the various linear (DCT, CPT, LPT, SLT, WAT, WT and WPT) and non -linear
(ENOPV, ENOCA, Maxlift, Medlift and LWT) transforms are studied and analyzed.
Among all the transforms, the best transforms are identified for ECG compression at low
and hi gh PRD. These transforms are then further used to improved transform based ECG
compression techniques. The transform based ECG compression techniques generally
involve transformation, thresholding, quantization and encoding steps.
The above transform base d technique is improved by introducing the normalization
process in between the transformation and thresholding. Here it is shown that the
normalization improves the performance of the transform based ECG compression
method. This improvement is more effect ive for high PRD. For example, the record MIT –
BIH 121 at UPRD=0.5, CR is 17.85 for DCT and at UPRD=2, CR is 95.65 for LWT

140
without normalization and with normalization CR increases to 20.85 for DCT and it
increases 148.12 for LWT.
In this thesis, alternativ e technique for accurate threshold calculation is also proposed and
analyzed. This technique is based on GA and it is shown that when GA is used the actual
PRD is almost equal to UPRD and therefore it is very effective when CR is performed
with low PRD. Th e computation time is less in case of low PRD (0.5) as compared to high
PRD (3).
Next, lifting schemes (Maxlift, Medlift, LWT) based quality controlled ECG compression
techniques are proposed and analyzed. These are non -linear transform based techniques.
Among the lifting techniques, LWT performs better. The LWT based technique is also
better than the linear transform based techniques for high PRD.
To avoid the problem of large number of coefficients at discontinuities, ENO interpolation
(ENOPV, ENOCA) base d ECG compression method is proposed where the number of
coefficients at discontinuities reduces to a lesser number which improves the compression
ratio. From the results it is observed that ENO interpolation based technique is a better
option over linear transforms based technique for high PRD.
In brief it can be stated that this thesis presents linear and non linear transform based
tunable or quality controlled ECG compression techniques. It is shown that one can
choose DCT, CPT and WAT or LWT depending o n the desirable quality. For signal
compression with high user PRD one can choose LWT and for low PRD DCT, CPT and
WAT can be selected.
The novel contributions of the research work presented in thesis are use of the
following to develop quality controlled ECG compression method with improved CR.
Use of Wave Atom Transform (WAT) based ECG compression method which
improves the CR at low PRD.

141

Use of Essentially Non Oscillatory (ENO) techniques to avoid discontinuities in
the ECG signal which leads to lesser n umber of coefficients at the edges and thus
improves the CR.
Use of Lifting wavelet transform (LWT) based ECG compression method which
provides better CR at high PRD.
The concept of normalization before thresholding which substantially increases the
CR.
Use of Genetic Algorithm (GA) to calculate the optimum value of threshold.
It is expected that the work presented in this thesis will be helpful in the design of an
efficient quality controlled ECG compression system.
5.2 Scope for Future Work
The work presente d in thesis can be extended to:
To design better ECG signal coding to improve the ECG signal compression.
To explore better quantization techniques for improvement in CR for ECG signal.
To explore better design for hardware implementation of ECG compressio n
techniques.
More ECG records can be selected for testing. In this thesis all the ECG are
original sampled at 360 Hz and digitized in 11 bits, future research can test ECG
records with different original sampling rate and different resolution levels.
Focu s on real -time applications, where the delays incurred in the encoding and
decoding processes should be minimal.

142

To explore the applications of ECG compression techniques for other medical
signals.
Work on methods other than transform based methods can be further explored.