Lukás Smital

Adaptive Wavelet Wiener Filtering of ECG Signals

In this study, we focused on the reduction of broadband myopotentials (EMG) in ECG signals using the wavelet Wiener filtering with noise-free signal estimation. We used the dyadic stationary wavelet transform (SWT) in the Wiener filter as well as in estimating the noise-free signal. Our goal was to find a suitable filter bank and to choose other parameters of the Wiener filter with respect to the signal-to-noise ratio (SNR) obtained. Testing was performed on artificially noised signals from the standard CSE database sampled at 500 Hz. When creating an artificial interference, we started from the generated white Gaussian noise, whose power spectrum was modified according to a model of the power spectrum of an EMG signal. To improve the filtering performance, we used adaptive setting parameters of filtering according to the level of interference in the input signal. We were able to increase the average SNR of the whole test database by about 10.6 dB. The proposed algorithm provides better results than the classic wavelet Wiener filter.

ECG features and methods for automatic classification of ventricular premature and ischemic heartbeats: A comprehensive experimental study

Accurate detection of cardiac pathological events is an important part of electrocardiogram (ECG) evaluation and subsequent correct treatment of the patient. The paper introduces the results of a complex study, where various aspects of automatic classification of various heartbeat types have been addressed. Particularly, non-ischemic, ischemic (of two different grades) and subsequent ventricular premature beats were classified in this combination for the first time. ECGs recorded in rabbit isolated hearts under non-ischemic and ischemic conditions were used for analysis. Various morphological and spectral features (both commonly used and newly proposed) as well as classification models were tested on the same data set. It was found that: a) morphological features are generally more suitable than spectral ones; b) successful results (accuracy up to 98.3% and 96.2% for morphological and spectral features, respectively) can be achieved using features calculated without time-consuming delineation of QRS-T segment; c) use of reduced number of features (3 to 14 features) for model training allows achieving similar or even better performance as compared to the whole feature sets (10 to 29 features); d) k-nearest neighbours and support vector machine seem to be the most appropriate models (accuracy up to 98.6% and 93.5%, respectively).

Towards real-time QRS feature extraction for wearable monitors

The ability to generate computationally compact ECG analysis algorithms is of interest in the field of wearable physiologic monitors. Such remote monitors necessarily have limited on-board energy storage and therefore lack the computational power and physical memory often required for academic study of physiologic waveforms. Herein we evaluate a set of algorithms with markedly different computation and memory footprints useful in extracting QRS complexes from synthetically generated noisy and measured ECG signals. A small memory and computational footprint Short Time Fourier Transform ECG analysis algorithm is demonstrated to have similar sensitivity and specificity to a more complex but highly accurate Stockwell Transform.

Automatic Detection of P Wave in ECG During Ventricular Extrasystoles

This work introduces a new method for P wave detection in ECG signals during ventricular extrasystoles. The authors of previous works which deal with detection of P waves tested their algorithms mainly on physiological records (sinus rhythm) and they reached good results for these records. Testing of P wave detection algorithms using pathological records is usually not provided and if it is, the results are notably worse than in the case of physiological records. The automatic and reliable detection of atrial activity in pathological situations is still an unsolved problem. In this work, phasor transform in combination with classification algorithm is used for P wave detection. Phasor transform converts each ECG sample into a phasor which enhances changes in the ECG signal. The classification is based on extraction of morphological features which are derived from each QRS complex. The results of classification are used for demarcation of areas in which P waves are searched using phasor transform. The proposed algorithm was tested on signals no. 106, 119, 214 and 223 from MIT-BIH arrhythmia database, in which the ventricular extrasystoles are present. For validation whether the algorithm is functional also for signals with physiological rhythm, it was tested on the signals no. 100, 101, 103, 117, and 122. The accuracy of the P wave detection in signals with ventricular extrasystoles is Se = 98.94% and PP = 98.30% and in signals without pathology is Se = 98.47% and PP = 99.99%.

Assessment of ECG Signal Quality After Compression

Highly efficient lossy compression algorithms for ECG signals are connected with distortion of the signals; lossy compression is a compromise between compression efficiency and signal quality. It is recommended to express this relation using rate-distortion curve. To decide whether the signal is suitable for further analysis, it is necessary to assess its quality after reconstruction. Although there exist many methods for quality assessment, neither of them is standardized or unified. The methods usually do not offer any information about their acceptable values. This paper introduces 10 new methods for signal quality assessment and their limits. Four methods are simple (entropy, mean, median, spectra similarity), two are based on delineation of ECG (SiP, SiPA), and four combine dynamic time warping, delineation, and calculation of distance (DTWdist, DTWpmfp1, DTWpmfp2, pmfp). These methods are tested on the whole standard CSE database using compression algorithm based on wavelet transform and set partitioning in hierarchical trees. The signals were compressed with various efficiency expressed by average value length (avL). Two ECG experts divided the compressed signals into three quality groups: perfect quality, good quality, not evaluable ECG. Owing to the experts’ ECG classification, we set the range of avL for each quality group. Based on this, we determined corresponding ranges of new methods’ values. Based on the trend of rate-distortion curve, its sensitivity, variability, their ratio at important boundary avL = 0.8 bps, and computational demand of the methods, we recommend four methods for further use.

Efficient implementation of Stockwell Transform for real-time embedded processing of physiologic signals

Physiologic monitoring enables scientists and physicians to study both normal and pathologic signals of the body. While wearable technologies are available today, many of these technologies are limited to data collection only. Embedded processors have minimal computational capabilities. We propose an efficient implementation of the Stockwell Transform which can enable real-time time-frequency analysis of biological signals in a microcontroller. The method is built upon the fact that the Stockwell Transform can be implemented as a compact filter bank with pre-computed filter taps. Additionally, due to the long tails of the gaussian windowing function, low amplitude filter taps can be removed. The method was implemented on a TI MSP430 processor. Simulated ECG data was fed into the processor to demonstrate performance and evaluate computational efficiency.

Optimization of the wavelet Wiener filtering for ECG signals

This paper deals with the methods of ECG signals denoising via wavelet Wiener filtering. We have studied the influence of the input parameters setting on filtered signals in a consideration of achieved signal-to-noise ratio (SNR). The Wiener filtering is used in the shift-invariant dyadic discrete-time wavelet domain for suppression of a parasite electromyographic (EMG) signal. To improve the filtering performance we used the adaptive adjustment of the method parameters, according to the level of the input noise. We are able to increase the average SNR of the whole tested database almost about 10 dB. The proposed algorithm provides better results, than a classic wavelet Wiener filtering method. The algorithm was tested on signals from the standard multilead CSE database.