Options
Wavelet transform in electrocardiograms
Author
Tay, Yian Ling
Supervisor
Kwek, Leong Chuan
Abstract
Understanding the heart’s structure, function and its electrical conduction system will enable us know more about our body and the existence of various heart diseases. An electrocardiogram (ECG) is used in diagnosing whether a heart is healthy or not. However, the interpretation of the ECG is complex and may vary slightly among different medical personnel.
ECG signals are measured as a function of time. More information about these signals can be obtained through signal analysis by transforming time domain to frequency domain. Fourier transform is commonly used to do so. A brief study on the various types of Fourier transform is done. Short time Fourier transform (STFT) overcomes the limitation of the Fourier transform as it is able to retrieve both frequency and time information from ECG signals. However, wavelet transform appears to be even more effective than STFT due to its capability to reveal concurrently the local spectral and temporal information from a signal in a more flexible way than STFT by utilising a window of variable width. Furthermore, wavelet transform analyses different frequencies with different resolutions.
The application of wavelet transform in ECGs has been studied by many people in recent years. This is because wavelet transform is very useful in signal coding, allowing information within the signal to be localised within a number of pertinent coefficients for compression purposes. The Haar wavelet transform, a type of discrete wavelet transform (DWT), uses filter banks for analysis and reconstructs time signals. It is also easy to implement and reduces the computation time and resources required.
The objective of this project is to assess the effectiveness of using Haar wavelet transform in ECG for diagnosis of IHD. Haar wavelet transform is used together with Donoho’s denoising method to analyse 64 data points from lead aVF in each ECG from four normal healthy patients and five ischaemic heart disease (IHD) patients. The common peaks in normal patients are identified and then used to compare with the peaks in IHD patients. It is shown that IHD patients have no peak at 0.20s, a timing which can be used to differentiate normal and IHD patients.
Due to the limited sample sizes for normal healthy patients and IHD patients, the results of this study lack specificity and sensitivity. Using larger sample sizes and exploring other leads in ECGs can be some areas to look into for future work to improve the validity and reliability of this project.
ECG signals are measured as a function of time. More information about these signals can be obtained through signal analysis by transforming time domain to frequency domain. Fourier transform is commonly used to do so. A brief study on the various types of Fourier transform is done. Short time Fourier transform (STFT) overcomes the limitation of the Fourier transform as it is able to retrieve both frequency and time information from ECG signals. However, wavelet transform appears to be even more effective than STFT due to its capability to reveal concurrently the local spectral and temporal information from a signal in a more flexible way than STFT by utilising a window of variable width. Furthermore, wavelet transform analyses different frequencies with different resolutions.
The application of wavelet transform in ECGs has been studied by many people in recent years. This is because wavelet transform is very useful in signal coding, allowing information within the signal to be localised within a number of pertinent coefficients for compression purposes. The Haar wavelet transform, a type of discrete wavelet transform (DWT), uses filter banks for analysis and reconstructs time signals. It is also easy to implement and reduces the computation time and resources required.
The objective of this project is to assess the effectiveness of using Haar wavelet transform in ECG for diagnosis of IHD. Haar wavelet transform is used together with Donoho’s denoising method to analyse 64 data points from lead aVF in each ECG from four normal healthy patients and five ischaemic heart disease (IHD) patients. The common peaks in normal patients are identified and then used to compare with the peaks in IHD patients. It is shown that IHD patients have no peak at 0.20s, a timing which can be used to differentiate normal and IHD patients.
Due to the limited sample sizes for normal healthy patients and IHD patients, the results of this study lack specificity and sensitivity. Using larger sample sizes and exploring other leads in ECGs can be some areas to look into for future work to improve the validity and reliability of this project.
Date Issued
2010
Call Number
RC683.5.E5 Tay
Date Submitted
2010