As a part of 2016 Physionet/CinC callenge, this work aims at the detection of abnormal phonocardiogram (PCG) recordings. Heart sound signal analysis has been an active research topic over the past decades with various studies such as heart sound segmentation and classification. We used the Physionet/CinC2016 challenge PCG database, which contains a large public collection of PCG recordings from a variety of clinical and nonclinical environments. The PCG classification in this work is performed in two steps. PCG heartbeats are first segmented and various heart sound markers are delineated. Then, a series of beat-specific features are extracted from the segmented heartbeats. Finally, PCG recordings are classified into normal and abnormal groups by performing classification based on tape-long features and by analyzing beat-extracted features from the PCG. Our method achieved an overall score of 80 in the unofficial phase of the challenge. In the official phase, the overall score of the proposed method was 82, with a sensitivity of 89%.

Introduction: An ECG signal has important information that can help for reflecting cardiac activity of a patient and medical diagnosis. Consistent or periodical heart rhythm disorders can result cardiac arrhythmias so classification algorithm for recognizing arrhythmias with satisfactory accuracy is necessary.Aims: In this study, a robust wavelet based algorithm for detection and delineation of events in ECG signal is applied and then a new synthesis of MLP-BP and PNN neural networks for heart arrhythmia classification was described.Methods: As a matter of fact any changes in the morphology of an ECG due to the arrhythmia are observed in time and frequency analysis so multi resolution analysis is applied for feature detection. First, noise and artifact is rejected by a discrete wavelet transform (DWT) and multi lead ECG is obtained. Then QRS complexes of signal is extracted and the signal is decomposed so corresponding DWT scales are segmented Next curve length and high order moment order based feature extraction are calculated for each excerpted segment and elements of feature vector for regulating the parameters of classifiers are obtained After generation of feature source and segmentation, Multi-Layer Perceptron-Back Propagation (MLP-BP) neural networks, Probabilistic Neural Network (PNN) and support vector machine (SVM) were designed and tuned and their results were compared.Results: The proposed algorithm was tested to all 48 record of the MIT-BIH arrhythmia database and also the proposed topology of classifiers and its related parameters is optimized by searching of best value of parameters. The average value of accuracy of each classifier over all records of MIT-BIH for arrhythmias recognition is Acc = 97.42, Acc = 98.24 and Acc = 97.42 for SVM, MLP and PNN classifiers respectively and also obtained results were compared with similar peer-reviewed studies in this subject.

The purpose of this study is to provide a new method for detecting fetal QRS complexes from non-invasive fetal electrocardiogram (fECG) signal. Despite most of the current fECG processing methods which are based on separation of fECG from maternal ECG (mECG), in this study, fetal heart rate (FHR) can be extracted with high accuracy without separation of fECG from mECG. Furthermore, in this new approach thoracic channels are not necessary. These two aspects have reduced the required computational operations. Consequently, the proposed approach can be efficiently applied to different real-time healthcare and medical devices. In this work, a new method is presented for selecting the best channel which carries strongest fECG. Each channel is scored based on two criteria of noise distribution and good fetal heartbeat visibility. Another important aspect of this study is the simultaneous and combinatorial use of available fECG channels via the priority given by their scores. A combination of geometric features and wavelet-based techniques was adopted to extract FHR. Based on fetal geometric features, fECG signals were divided into three categories, and different strategies were employed to analyze each category. The method was validated using three datasets including Noninvasive fetal ECG database, DaISy and PhysioNet/Computing in Cardiology Challenge 2013. Finally, the obtained results were compared with other studies. The adopted strategies such as multi-resolution analysis, not separating fECG and mECG, intelligent channels scoring and using them simultaneously are the factors that caused the promising performance of the method.

The most straightforward method for heart beat estimation is R-peak detection based on an electrocardiogram (ECG) signal. Current R-peak detection methods do not work properly when the ECG signal is contaminated or missing, which leads to the incorrect estimation of the heart rate. This raises the need for reliable algorithms which can locate heart beats in continuous long-term multimodal data, allowing robust analysis.In this paper, three peak detectors are evaluated for heart beat detection using various cardiovascular signals. One of the peak detectors is a new general peak detector (GPD) algorithm which is applicable on ECG and other pulsatile signals to compensate for the limitation of QRS detection. This peak detector algorithm is adaptive and independently finds amplitude characteristics for every recording, while not tuned for ECG or other pulsatile signals. Three strategies, which are different disciplines of detectors, are then proposed while the fusion method remains the same in all strategies. In the first strategy, the ECG and the lowest-indexed signal of general blood pressure (BP), arterial blood pressure (ART) and pulmonary arterial pressure (PAP) are processed through gqrs and wabp (from the PhysioNet library), respectively. In the second strategy, all beats in different signals are detected by GPD. In the third strategy, ECG and other signals are processed by gqrs and GPD, respectively. In all three strategies two criteria are used in order to fuse the detections. The first criterion is based on the number of candidate detections in a specific time period, based on which signals of interest are selected. The second fusion criterion is based on the regularity of the derived intervals between subsequent candidate detections. If the number of detections in ECG and one of BP, ART and PAP signals have reasonable physiological range, a new signal is generated in which they are coupled with each other. Heart beats can more easily be detected in noisy parts of these signals using the new coupled waveform. For instance, if ECG and BP are coupled, BP pulses make the real heart beats in noisy parts of ECG detectable and ECG R-peaks make the weak BP pulses detectable in the new waveform. The proposed peak detector is developed using the MIT/BIH arrhythmia database. Furthermore, heart beat detection strategies were evaluated using the train and test datasets of PhysioNet/CinC Challenge (2014), and the overall results of the strategies are compared.

Introduction: Coronary artery congestion is a heart disease which causes a lack of oxygen and nutrients in the heart, and is felt as chest pain (ischemia disease). Prolonged ischemia can continue until the cells start to dye, which is called myocardial infraction.Aims: We aim to determine the amount of cardiac tissue damage by multi resolution analysis of electrocardiogram (ECG) and atrial blood pressure (ABP) signals.Methods: In this study 39 Wistar rats were used, and ECG and ABP signals were recorded to estimate the extent of tissue damage. The signals were recorded for 30 minutes during normal heart function and for 30 minutes during ischemia and myocardial infraction (MI) which was induced by artificial complete blockage of the left anterior descending coronary artery (LAD). Additionally the vasopressin (AVP) in various doses was injected to 39 rats. Afterwards, the wavelet packet transform (WPT) wad applied to the recorded ECG and ABP signals for decomposition into dyadic scales. 50 dyadic scales which were more informative for discrimination of different ischemic phases among each other and from the healthy phase were chosen, and a feature vector as the entropy of corresponding wavelet coefficients was selected. As a reference measurement of the extent of damage tissue, images of the heart sections were additionally extracted, and the extent of damage tissue was assessed by image processing technique. Finally, the amount of damage tissue was classified with artificial neural networks (ANN) based techniques.Results: The extent of tissue damage was estimated based on the ANN and multi-resolution analysis of the synchronic electromechanical signals with the average error of the 2.17% for the normal and ischemic tissue in all the AVP doses.

The aim of this study is the intelligent recognition of the fetal heart rate and its R-R intervals from noninvasive fetal electrocardiogram signals. The non-value data was first eliminated and the missing data were regenerated based on the statistical distribution of the data. Then, the power line noise and baseline noise are removed. At the next step, a variable threshold criterion was designed to detect the maternal R-waves. By eliminating the specific ranges of the maternal R waves from signal, the remaining data describe merely the fetal QRS complexes. Next, a window with a specific length was slid on D1 signals and the envelope curves were extracted. The locations of each local maximum on the envelope curve represent the fetal R waves. Finally, in order to improve both the performance of the proposed method and the robustness of the algorithm to noise, an amendment technique with respect to the fetal and maternal R-R intervals was implemented The algorithm was applied on the test data set B consequently as the preliminary challenge scores. The average scores 108.766 and 15.480 were achieved as the best scores for the events 4 and 5, respectively, on phase 1, and 63.750 and 11.198 on phase 2.

In this study, a new supervised noise-artifact-robust heart arrhythmia fusion classification solution, is introduced. Proposed method consists of structurally diverse classifiers with a new QRS complex geometrical feature extraction technique. Toward this objective, first, the events of the electrocardiogram (ECG) signal are detected and delineated using a robust wavelet-based algorithm. Then, each QRS region and also its corresponding discrete wavelet transform (DWT) are supposed as virtual images and each of them is divided into eight polar sectors. Next, the curve length of each excerpted segment is calculated and is used as the element of the feature space. Discrimination power of proposed classifier in isolation of different Gold standard beats was assessed with accuracy 98.20%. Also, proposed learning machine was applied to 7 arrhythmias belonging to 15 different records and accuracy 98.06% was achieved. Comparisons with peer-reviewed studies prove a marginal progress in computerized heart arrhythmia recognition technologies.

This paper describes a new supervised fusion (hybrid) electrocardiogram (ECG) classification solution consisting of a new QRS complex geometrical feature extraction as well as a new version of the learning vector quantization (LVQ) classification algorithm aimed for overcoming the stability-plasticity dilemma. Toward this objective, after detection and delineation of the major events of ECG signal via an appropriate algorithm, each QRS region and also its corresponding discrete wavelet transform (DWT) are supposed as virtual images and each of them is divided into eight polar sectors. Then, the curve length of each excerpted segment is calculated and is used as the element of the feature space. To increase the robustness of the proposed classification algorithm versus noise, artifacts and arrhythmic outliers, a fusion structure consisting of five different classifiers namely as Support Vector Machine (SVM), Modified Learning Vector Quantization (MLVQ) and three Multi Layer Perceptron-Back Propagation (MLP–BP) neural networks with different topologies were designed and implemented. The new proposed algorithm was applied to all 48 MIT–BIH Arrhythmia Database records (within–record analysis) and the discrimination power of the classifier in isolation of different beat types of each record was assessed and as the result, the average accuracy value Acc=98.51% was obtained. Also, the proposed method was applied to 6 number of arrhythmias (Normal, LBBB, RBBB, PVC, APB, PB) belonging to 20 different records of the aforementioned database (between– record analysis) and the average value of Acc=95.6% was achieved. To evaluate performance quality of the new proposed hybrid learning machine, the obtained results were compared with similar peer– reviewed studies in this area. Keywords—Feature Extraction, Curve Length Method, Support Vector Machine, Learning Vector Quantization, Multi Layer Perceptron, Fusion (Hybrid) Classification, Arrhythmia Classification, Supervised Learning Machine.

The aim of this study is to describe a new false-alarm probability (FAP) bounded unified framework for segmentation of the phonocardiogram (PCG) signal sounds registered by an electronic stethoscope board. To meet this end, first the original PCG signal is pre-processed by application of an appropriate bandpass finite-duration impulse response (FIR) filter and then by implementation of a trous discrete wavelet transform (DWT) to the filtered signal for extracting several dyadic scales. Then, after choosing a proper scale, a fixed sample size sliding window is moved on the selected scale and in each slide, six feature vectors namely summation of the nonlinearly amplified Hilbert transform, summation of absolute first order differentiation, summation of absolute second-order differentiation, curve length, area and variance of the excerpted segment are calculated. Then, all feature trends are normalized and utilized to construct a newly proposed principal components analyzed geometric index (PCAGI) (to be used as the segmentation decision statistic (DS)) by application of a linear orthonormal projection. Next, using an adaptive smoothing filter (ASF), the obtained metric is modulated and freed from the fast fluctuations occurring in the vicinity of events onset and offset locations which consequently results in enhancement of edge detection accuracy. Later, histogram parameters of the filtered DS metric are used for the regulation of the alpha-level Neyman-Pearson classifier for FAP-bounded delineation of the PCG events. To assess the performance quality of the proposed PCG segmentation algorithm, the method was applied to all 85 records of Nursing Student Heart Sounds database including stenosis, insufficiency, regurgitation, gallop, septal defect, sound split, rumble, murmur, clicks, friction rub and snap disorders with different sampling frequencies. The method was also applied to the records obtained from an electronic stethoscope board designed for fulfillment of this study in the presence of high-level power-line noise and external disturbing sounds and as a result, no false positive or false negative errors were detected. High robustness against measurement noises of the electronic stethoscopes, acceptable detection-segmentation accuracy of PCG events in the presence of severe heart valvular and arrhythmic dysfunctions within a tolerable computational burden (processing time) and having no parameter dependency on the acquisition sampling frequency can be mentioned as important merits and capabilities of the proposed PCAGI-based PCG events detection-segmentation algorithm. Copyright (C) 2011 John Wiley & Sons, Ltd.

The paper addresses a new QRS complex geometrical feature extraction technique as well as its application for electrocardiogram (ECG) supervised hybrid (fusion) beat-type classification. To this end, after detection and delineation of the major events of ECG signal via a robust algorithm, each QRS region and also its corresponding discrete wavelet transform (DWT) are supposed as virtual images and each of them is divided into eight polar sectors. Then, the curve length of each excerpted segment is calculated and is used as the element of the feature space. To increase the robustness of the proposed classification algorithm versus noise, artifacts and arrhythmic outliers, a fusion structure consisting of three Multi Layer Perceptron-Back Propagation (MLP-BP) neural networks with different topologies and one Adaptive Network Fuzzy Inference System (ANFIS) were designed and implemented. To show the merit of the new proposed algorithm, it was applied to all MIT-BIH Arrhythmia Database records and the discrimination power of the classifier in isolation of different beat types of each record was assessed and as the result, the average accuracy value Acc=98.27% was obtained. Also, the proposed method was applied to 8 number of arrhythmias (Normal, LBBB, RBBB, PVC, APB, VE, PB, VF) belonging to 19 number of the aforementioned database and the average value of Acc=98.08% was achieved. To evaluate performance quality of the new proposed hybrid learning machine, the obtained results were compared with similar peer-reviewed studies in this area.

The paper addresses a new QRS complex, geometrical feature extraction technique, as well as its application in supervised electrocardiogram (ECG) heart-beat hybrid (fusion) classification. To this end, after detection and delineation of the major events of an ECG signal via an appropriate algorithm, each QRS region and also its corresponding Discrete Wavelet Transform (DWT) are supposed as virtual images, and each one is divided into eight polar sectors. Then, the curve length of each excerpted segment is calculated and used as an element of the feature space. To increase the robustness of the proposed classification algorithm versus noise, artifacts and arrhythmic outliers, a fusion structure consisting of four different classifiers, namely Support Vector Machine (SVM), Probabilistic Neural Network (PNN) and two Multi Layer Perceptron-Back Propagation (MLP-BP), with different topologies, were designed. To show the merit of the new proposed algorithm, it was applied to all MIT-BIH arrhythmia database records, and the discriminative power of the classifier in isolation of different beat types of each record was assessed. As a result, the average accuracy value, Acc=98.18%, was obtained. Also, the proposed method was applied to 8 arrhythmias and an average value of Acc=97.37% was achieved.

Since ambulatory electrocardiogram (ECG) signal is always accompanied by strong noise, high amplitude baseline wandering, impulsive artifacts, arrhythmic outliers and some discontinuities, these effects reduce the accuracy of a computerized cardiac-originated events detection-delineation algorithm. The aim of this study is to describe a multi-aspect robust structure of a solution designed for detection-delineation of major events of the long-duration holter ECG signal.

The aim of this study is to describe a robust unified framework for segmentation of the phonocardiogram (PCG) signal sounds based on the false-alarm probability (FAP) bounded segmentation of a properly calculated detection measure. To this end, first the original PCG signal is appropriately pre-processed and then, a fixed sample size sliding window is moved on the pre-processed signal. In each slid, the area under the excerpted segment is multiplied by its curve-length to generate the Area Curve Length (ACL) metric to be used as the segmentation decision statistic (DS). Afterwards, histogram parameters of the nonlinearly enhanced DS metric are used for regulation of the α-level Neyman-Pearson classifier for FAP-bounded delineation of the PCG events. The proposed method was applied to all 85 records of Nursing Student Heart Sounds database (NSHSDB) including stenosis, insufficiency, regurgitation, gallop, septal defect, split sound, rumble, murmur, clicks, friction rub and snap disorders with different sampling frequencies. Also, the method was applied to the records obtained from an electronic stethoscope board designed for fulfillment of this study in the presence of high-level power-line noise and external disturbing sounds and as the results, no false positive (FP) or false negative (FN) errors were detected. High noise robustness, acceptable detection-segmentation accuracy of PCG events in various cardiac system conditions, and having no parameters dependency to the acquisition sampling frequency can be mentioned as the principal virtues and abilities of the proposed ACL-based PCG events detection-segmentation algorithm.