Ana Gavrovska

Automatic heart sound detection in pediatric patients without electrocardiogram reference via pseudo-affine Wigner-Ville distribution and Haar wavelet lifting

Having in mind the availability of electronic stethoscopes, phonocardiograms (PCGs) have become popular for cardiovascular functionality monitoring and signal processing applications. Detection of fundamental heart sounds (HSs), S1s and S2s, is considered to be a crucial step in PCG analysis. Electrocardiogram (ECG), noted as a reference signal, is often synchronously recorded in order to simplify the S1/S2 detection process. Nevertheless, electronic stethoscopes are frequently used without additional ECG equipment. We propose a new algorithm for automatic fundamental HSs detection via: joint time-frequency representation based on pseudo affine Wigner-Ville distribution (PAWVD), Haar wavelet lifting scheme (Haar-LS), normalized average Shannon energy (NASE) and autocorrelation. The performance of the proposed algorithm was calculated on both normal (50) and pathological (75) PCG recordings, eight seconds long each, contributed by 125 different pediatric patients. The algorithm showed relatively high recall (90.41%) and precision (96.39%) rates of S1/S2 detection procedure in a variety of PCG signals, without ECG as a reference. Furthermore, it indicated the ability to overcome splitting within the S1/S2 heart sounds.

Classification of Prolapsed Mitral Valve versus Healthy Heart from Phonocardiograms by Multifractal Analysis

Phonocardiography has shown a great potential for developing low-cost computer-aided diagnosis systems for cardiovascular monitoring. So far, most of the work reported regarding cardiosignal analysis using multifractals is oriented towards heartbeat dynamics. This paper represents a step towards automatic detection of one of the most common pathological syndromes, so-called mitral valve prolapse (MVP), using phonocardiograms and multifractal analysis. Subtle features characteristic for MVP in phonocardiograms may be difficult to detect. The approach for revealing such features should be locally based rather than globally based. Nevertheless, if their appearances are specific and frequent, they can affect a multifractal spectrum. This has been the case in our experiment with the click syndrome. Totally, 117 pediatric phonocardiographic recordings (PCGs), 8 seconds long each, obtained from 117 patients were used for PMV automatic detection. We propose a two-step algorithm to distinguish PCGs that belong to children with healthy hearts and children with prolapsed mitral valves (PMVs). Obtained results show high accuracy of the method. We achieved 96.91% accuracy on the dataset (97 recordings). Additionally, 90% accuracy is achieved for the evaluation dataset (20 recordings). Content of the datasets is confirmed by the echocardiographic screening.

Selection of wavelet decomposition levels in ECG filtering

Optimal selection of decomposition levels in wavelet transform used for both HF and LF filtering of ECG signal is described. By analyzing different types of wavelets and selected decomposition levels, the best parameters are determined and their statistical performances are derived. Three Physionets ECG databases were used as a test-bed.

Application of wavelet and EMD-based denoising to phonocardiograms

Wavelet-based (WT) denoising has been frequently used as simple and efficient method enabling high signal-to-noise ratio. Moreover, by using second generation WT and lifting scheme, even better performance has been obtained. Besides the WT, empirical mode decomposition (EMD) is another efficient noise reduction technique, which enables adaptive decomposition based only on signal characteristics. Both WT and EMD denoising concepts find place in pre-processing of phonocardiograms (PCGs). Although these techniques are very effective in denoising, their inappropriate use can cause unwanted signal distortion. In this paper, we analyzed both denoising techniques and tested their influence to the degradation of PCGs, particularly regarding to some diagnostically relevant cardiac events, such as the first and second heart sound (S1 and S2) and clicks. It was shown that the morphology of S1, S2, and even the existence of subtle features (clicks) within the PCG, depend on the denoising method. Results of evaluation tests indicate which combination of denoising algorithm is the best choice for PCG signal pre-processing.

Region-based phonocardiogram event segmentation in spectrogram image

Assisting physicians in the auscultation of the patients by providing cost effective, automatic and accurate signal processing module for the heart event detection and recognition is of importance. We suggest that the segmentation of time-frequency representation image of one-dimensional phonocardiogram signals (PCGs) could be effective for observing possible abnormal heart events. The fact that frequency bands of cardiac events overlap and that different heart dysfunctions can occur simultaneously make image event segmentation a difficult task. Ambient noise and artifact burden make the segmentation problem even more difficult. A method of “splitting and merging” algorithm involving region-growing based on seeds is presented. We demonstrate that this method contributes to design of patterns required for cardiac event recognition and identification on several examples presented here.

Second generation wavelets: Advantages in cardiosignal processing

In this paper, the cardiosignal analysis using second generation wavelets and lifting structure based on clinically-relevant extracted information is described. The information extraction can be performed by different spectral (such as multifractal) and spectrogram (time-frequency) representations. Here, the novelties are related mainly to the knowledge-controlled analysis and noise reduction in phonocardiograms. We suggest adaptive (signal-dependent) wavelet-based processing able to simultaneously preserve relevant high-frequency details and reduce noise.

Face recognition using Gabor filters, PCA and neural networks

This paper proposes a face recognition system based on Gabor image representation, principal component analysis and neural networks. Gabor filter bank at 5 scales and 8 orientations is used for face representation. The dimensionality of Gabor representation of face images is reduced by the Principal component analysis. The images are classified using the Feedforward neural network with backpropagation learning algorithm. The ORL database of faces is used for testing.

PVC scalogram detection using neural network

This paper defines pattern matching using backpropagation neural network for detection of Premature Ventricular Complexes in electrocardiograms. Scalograms using continuous wavelet transformation proved to be a powerful tool in morphological recognition of the segments. Total segment overlapping is also presented as a convenient method for classifying normal and odd intervals. It can be helpful not only in interval segmentation for making patterns, but also as a visualization tool.

Computer-based analysis of cardiac sounds based on multiscale techniques

In this paper special attention is given to domain selection and local and global signal examination which is important for signal processing, especially in cardiosignals where physiological information is a priority. In the case of phonocardiographic signals, spectral components are analyzed using joint time-frequency and multifractal representation, as well as directly in time domain using second generation wavelets and normalized averaged Shannon energy. This paper describes novel techniques for computer-based analysis of heart sounds without reference electrocardiograms. Experimental performance results of multiscale based model show excellent results in fundamental heart sound detection and abnormality indication via singularity reoccurrence detection with preserving relevant singularities.

Potential applications of HDR in echocardiography

The diagnostic process in medicine in many cases depends on readability of medical images. Thus, the image enhancement is very important step in medical imaging. We investigate the potential application of High Dynamic Range (HDR) imaging in echocardiography. HDR image is generated using two Low Dynamic Range (LDR) echocardiograms. Since many devices can display only limited range of luminous intensity, the dynamic range of input HDR image needs to be reduced. This is done by operation called tone mapping. Several HDR tone mapping methods are analyzed and the results are presented in the paper.