Jiří Kozumplík

Wavelet transform in electrocardiography—data compression

An application of the wavelet transform to electrocardiography is described in the paper. The transform is used as a first stage of a lossy compression algorithm for efficient coding of rest ECG signals. The proposed technique is based on the decomposition of the ECG signal into a set of basic functions covering the time-frequency domain. Thus, non-stationary character of ECG data is considered. Some of the time-frequency signal components are removed because of their low influence to signal characteristics. Resulting components are efficiently coded by quantization, composition into a sequence of coefficients and compression by a run-length coder and a entropic Huffman coder. The proposed wavelet-based compression algorithm can compress data to average code length about 1 bit/sample. The algorithm can be also implemented to a real-time processing system when wavelet transform is computed by fast linear filters described in the paper.

A Wavelet-Based ECG Delineation in Multilead ECG Signals: Evaluation on the CSE Database

In this paper, we present simple and fast approach for electrocardiogram (ECG) delineation based on a continuous wavelet transform (CWT). First of all, QRS complexes are detected. Then, QRS onset and end are found for each QRS complex. In the next step, the T wave is detected between the QRS end and the next QRS onset. The T wave end is found between the position of T wave and the next QRS onset. Finally, the P wave is detected between the T wave end and the next QRS onset. In the last step, the P wave onset and end are found. These significant points are found separately in each lead of the ECG signal. Global multilead positions were determined from singlelead positions by using special selection rule. The presented algorithm was evaluated on the standard CSE database, which contains reference global positions common for all leads. We obtained two sets of results, first for evaluation on 12 standard leads and second for evaluation on Frank leads. Calculated standard deviations for 12 standard leads were much smaller than given accepted tolerances for most of the characteristic points and standard deviations for Frank leads accomplished given tolerances for most of the characteristic points. The proposed algorithm did well in comparison with other algorithms, especially standard deviations for the T wave end are very interesting: s = 12.2 ms (12 standard leads), s = 19.7 ms (Frank leads).

Heart rate variability analysed by Poincaré plot in patients with metabolic syndrome

INTRODUCTION The SD1 and SD2 indexes (standard deviations in two orthogonal directions of the Poincaré plot) carry similar information to the spectral density power of the high and low frequency bands but have the advantage of easier calculation and lesser stationarity dependence. METHODS ECG signals from metabolic syndrome (MetS) and control group patients during tilt table test under controlled breathing (20 breaths/minute) were obtained. SD1, SD2, SDRR (standard deviation of RR intervals) and RMSSD (root mean square of successive differences of RR intervals) were evaluated for 31 control group and 33 MetS subjects. RESULTS Statistically significant lower values were observed in MetS patients in supine position (SD1: p=0.03, SD2: p=0.002, SDRR: p=0.006, RMSSD: p=0.01) and during tilt (SD2: p=0.004, SDRR: p=0.007). CONCLUSION SD1 and SD2 combining the advantages of time and frequency domain methods, distinguish successfully between MetS and control subjects.

Changes in time-frequency phase spectra vs. ST-segment deviation for detecting acute coronary artery occlusion

The paper deals with a new method for the detection of myocardial ischemia caused by acute coronary artery occlusion, in the early phases. The method is based on the analysis of intra-QRS changes in time-frequency phase spectra generated by a wavelet transform. To verify the method, 11 Langendorff-perfused rabbit hearts have been used. The presented results show that models can detect early ischemia in one of three orthogonal leads as early as one minute after coronary artery occlusion. In addition, the method is compared to the traditional ST-segment analysis.

Time — Varying filters for ECG processing

The paper deals with simple approach to digital time-varying ECG filtering. The theoretical presuppositions of the filter performance are described as well as the method of the filter design.