Heart sound classification based on sub-band envelope and convolution neural network / 生物医学工程学杂志

Automatic classification of heart sounds plays an important role in the early diagnosis of congenital heart disease. A kind of heart sound classification algorithms based on sub-band envelope feature and convolution neural network was proposed in this paper, which did not need to segment the heart sounds according to cardiac cycle accurately. Firstly, the heart sound signal was divided into some frames. Then, the frame level heart sound signal was filtered with Gammatone filter bank to obtain the sub-band signals. Next, the sub-band envelope was extracted by Hilbert transform. After that, the sub-band envelope was stacked into a feature map. Finally, type Ⅰ and type Ⅱ convolution neural network were selected as classifier. The result shown that the sub-band envelope feature was better in type Ⅰ than type Ⅱ. The algorithm is tested with 1 000 heart sound samples. The test results show that the overall performance of the algorithm proposed in this paper is significantly improved compared with other similar algorithms, which provides a new method for automatic classification of congenital heart disease, and speeds up the process of automatic classification of heart sounds applied to the actual screening.

An inversion of the conical Radon transform arising in the Compton camera with helical movement

Since the Compton camera was fi rst introduced, various types of conical Radon transforms have been examined. Here, we derive the inversion formula for the conical Radon transform, where the cone of integration moves along a curve in three-dimensional space such as a helix. Along this three-dimensional curve, a detailed inversion formula for helical movement will be treated for Compton imaging in this paper. The inversion formula includes Hilbert transform and Radon transform. For the inversion of Compton imaging with helical movement, it is necessary to invert Hilbert transform with respect to the inner product between the vertex and the central axis of the cone of the Compton camera. However, the inner product function is not monotone. Thus, we should replace the Hilbert transform by the Riemann–Stieltjes integral over a certain monotone function related with the inner product function. We represent the Riemann–Stieltjes integral as a conventional Riemann integral over a countable union of disjoint intervals, whose end points can be computed using the Newton method. For the inversion of Radon transform, three dimensional fi ltered backprojection is used. For the numerical implementation, we analytically compute the Hilbert transform and Radon transform of the characteristic function of fi nite balls. Numerical test is given, when the density function is given by a characteristic function of a ball or three overlapping balls.

Detección Automática y Clasificación de Eventos en Sonidos Cardiopulmonares de Sujetos Saludables / Automatic Detection and Classification of Cardiopulmonary Events

Resumen: La auscultación de señales basada en un estetoscopio estándar y/o electrónico no solo incluye sonidos internos del cuerpo, también incluye frecuentemente ruido externo de interferencia con componentes en el mismo rango. Esta forma de examinar es incluso afectada por los umbrales auditivos variantes de los profesionales de la salud y el grado de experiencia en reconocimiento de indicadores peculiares. Además, los resultados son a menudo caracterizados en términos cualitativos descriptivos sujetos a interpretaciones individuales. Para direccionar esta preocupación, los estudios presentados en este artículo contienen un procesamiento concurrente de las componentes dominantes de sonidos del corazón (HS) y del pulmón (HS), y una etapa de acondicionamiento que incluye la reducción de HS presente en señales LS. Específicamente, la transformada de Hilbert fue una técnica de caracterización para HS. En el caso de señales enfocadas a LS, las técnicas de detección de actividad de voz y el cálculo de umbrales de algunos componentes de los vectores acústicos de Coeficientes Cepstrales en Frecuencia Mel (MFCC), fueron útiles en la caracterización de eventos acústicos asociados. Las fases de inspiración y expiración fueron diferenciadas por medio de la sexta componente de MFCC. Con el fin de evaluar la eficiencia de esta aproximación, proponemos los Modelos Ocultos de Markov con Modelos Mesclados Gaussianos (HMM-GMM). Los resultados utilizando esta forma de detección son superiores cuando se desarrolla la clasificación con modelos HMM-GMM, la cual refleja las ventajas de la forma de detección cuantificable y clasificación sobre la aproximación clínica tradicional.

Abstract: A standard and/or electronic stethoscope based auscultatory signals include not only the internal sounds of the body but also interfering external noise often with similar frequency components. This form of examination is also affected by varying thresholds of clinical practitioner's hearing and degree of experience in recognition of peculiar auscultatory indicators. Further, the results are often characterized in qualitative descriptive terms subject to individual's interpretation. To address these concerns, presented studies include concurrent processing of dominant heart (HS) and lung (LS) sounds components and a conditioning stage involving HS presence reduction within LS focused signals. Specifically as determined, the Hilbert transform was a technique of choice in HS characterization. In the case of LS focused signals, the speech activity detection techniques (VAD) and the thresholds calculation of some components of acoustic vectors of Cepstral Coefficients in Mel Frequency (MFCC), were useful in characterization of associated acoustic events. The phases of inspiration and expiration were differentiated by means of the sixth component of MFCC. In order to evaluate the efficiency of this approach, we propose Hidden Markov Models with Mixed Gaussian Models (HMM-GMM). The results utilizing this form of detection are superior when performing classification with HMM-GMM models, which reflect the advantages of presented form of quantifiable detection and classification over traditional clinical approach.

Modeling and FPGA-based implementation of an efficient and simple envelope detector using a Hilbert Transform FIR filter for ultrasound imaging applications

Abstract Introduction Although the envelope detection is a widely used method in medical ultrasound (US) imaging to demodulate the amplitude of the received echo signal before any back-end processing, novel hardware-based approaches have been proposed for reducing its computational cost and complexity. In this paper, we present the modeling and FPGA implementation of an efficient envelope detector based on a Hilbert Transform (HT) approximation for US imaging applications. Method The proposed model exploits both the symmetry and the alternating zero-valued coefficients of a HT finite impulse response (FIR) filter to generate the in-phase and quadrature components that are necessary for the envelope computation. The hardware design was synthesized for a Stratix IV FPGA, by using the Simulink and the integrated DSP Builder toolbox, and implemented on a Terasic DE4-230 board. The accuracy of our algorithm was evaluated by the normalized root mean square error (NRMSE) cost function in comparison with the conventional method based on the absolute value of the discrete-time analytic signal via FFT. Results An excellent agreement was achieved between the theoretical simulations with the experimental result. The NRMSE was 0.42% and the overall FPGA utilization was less than 1.5%. Additionally, the proposed envelope detector is capable of generating envelope data at every FPGA clock cycle after 19 (0.48 µs) cycles of latency. Conclusion The presented results corroborate the simplicity, flexibility and efficiency of our model for generating US envelope data in real-time, while reducing the hardware cost by up to 75%.

Uma arquitetura para detecção online de transientes em sinais de eletrocardiograma sobre o protocolo PM-AH / An architecture for online transient detection in electrocardiogram signals on the MP-HA protocol

Uma identificação correta de transientes em sinais de ECG (Eletrocardiograma) pode auxiliar métodos de processamento de sinal de ECG, pois esse tipo de evento degrada o sinal e pode induzir a erros. Diante disso, o presente trabalho propõe uma arquitetura para a detecção desses fenômenos, seguindo a tendência atual da computação distribuída, na qual um sensor realiza a detecção dos transientes no momento da aquisição do sinal, e, em seguida, encaminha essa informação através de uma rede de comunicação de dados, desenvolvida especialmente para a automação hospitalar, até um dispositivo computacional que irá processar os dados ou então apresentá-los a um profissional capacitado para fazer a análise de forma manual. Para realizar a detecção de transientes, foi proposto um método matemático baseado na transformada Hilbert do sinal de ECG, aliado ao PM-AH (Protocolo Multiciclos para Automação Hospitalar), com adição de quadros neste, para que seja possível o envio da informação sobre a ocorrência de transientes junto aos dados do sinal de eletrocardiograma. Dentre os transientes possíveis, foi escolhido o ruído, por ser o fenômeno que mais interfere no processamento de sinais de ECG, onde testes foram realizados com a base de dados MIT-BIH Arrhythmia Database, enquanto uma análise matemática foi feita nos novos quadros do protocolo PM-AH, com o intuito de demonstrar a consistência do protocolo com esta adição.

A correct identification of transients in the ECG (electrocardiogram) can assist processing methods for ECG signals, since this type of event degrades the signal and can be misleading. Therefore, this paper proposes an architecture for detection of these phenomena, following the current trend of distributed computing, in which a sensor will detect transients at the time of signal acquisition, and then forward this information through a data communication network, designed specifically for hospital automation, to a computing device that will process the data or present it to a trained professional for manual analysis. To perform the detection of transients, a mathematical method based on the Hilbert transform of the ECG signal is proposed here, allied with the MP-HA (Multicycle Protocol for Hospital Automation), with the addition of frames, so that information on the occurrence of transients can be transmitted along with signal data of the electrocardiogram. Among the possible transients, noise was chosen because it is the phenomenon that interferes the most with the processing of ECG signals. Tests were performed using the MIT-BIH Arrhythmia Database, while a mathematical analysis was used in the new frames of the MP-HA protocol in order to demonstrate the consistency of the protocol with this addition.

Análise comparativa de desempenho das transformadas Wavelet e Hilbert na detecção do QRS em ECG / Performance comparison analysis of Wavelet and Hilbert transforms for QRS detection in ECG

O processo de detecção do complexo QRS é o primeiro passo de um processo de extração de parâmetros do sinal eletrocardiograma (ECG) em sistemas de auxílio ao diagnóstico médico. O presente trabalho apresenta resultados detalhados de comparação da aplicação de duas transformadas matemáticas, Wavelet e Hilbert, em um algoritmo de detecção de QRS em termos de taxas de detecções corretas (sensibilidade e preditividade positiva) e de uma medida de frequência de recorrência a processos de filtragem (pré-processamento). Uma abordagem inovadora é implementada, na qual as rotinas de filtragem são inseridas dentro do estágio de decisão, ou seja, é realizada a supressão da etapa de pré-processamento. As transformadas são aplicadas no algoritmo, que é baseado em um limiar adaptativo, com o objetivo de realçar, apenas quando necessário, os picos (pontos fiduciais)do QRS. Em uma primeira abordagem, apenas a transformada Wavelet é utilizada neste realce e, numa segunda abordagem, a transformada de Hilbert é inserida em série à aplicação da Wavelet em dois possíveis arranjos. São realizados experimentos dos algoritmos sobre os exames da base de dados Arrhythmia Database, pertencente ao conjunto de bases de dados do MIT-BIH. É composta por 48 gravações de ECG com duração de trinta minutos, amostrados a uma frequência de 360 Hz com resolução de 4,88 μV sobre uma faixa de variação de 10 mV. Ao todo, contabilizam-se 109.662 complexos QRS. Taxas de 98,85% de sensibilidade e 95,10% de preditividade positiva são obtidas com a aplicação exclusiva da transformada Wavelet, enquanto que 98,89% de sensibilidade e 98,52% de preditividade positiva são obtidas com aaplicação em série das transformadas Wavelet e de Hilbert.

The process of QRS detection is the first stage of a greater process: the feature extraction in the electrocardiogram (ECG). This work presents detailed results on the performance of two mathematical transforms, Hilbert and Wavelet, which are applied in QRS detection. The evaluation parameters are the detection rates and a measure of frequency of recurrence to filtering processes. An innovative approach is implemented: the filtering routines are inserted in the decision stage, i.e. the preprocessing stage is removed. The algorithm is based on adaptive threshold technique and the two transforms are applied in order to emphasize, only when necessary, the QRS fiducial points. In a first approach, only the Wavelet transform is applied, and in a second approach, the Hilbert transform is inserted before the Wavelet transform or after it. We evaluate these approaches on the well-known MIT-BIH Arrhythmia Database. It contains 48 half-hour recordings of annotated ECG with a sampling rate of 360 Hz and 4.88 μV resolution over a 10 mV range, totalizing 109,662 QRS complexes. Sensitivity rates of 98.85% and 98.89% are respectively attained when the Wavelet transform is applied in the filtering processes and both Hilbert and Wavelet transforms are applied. Predictability rates of 95.10% and 98.52% are also attained respectively using Wavelet transform and the simultaneous application of Hilbert and Wavelet transforms in the filtering processes.

A Novel Method for Extraction of Heart Sounds Envelope Based on Interpolation of Local Peak and Valley Points / 航天医学与医学工程

Objective To extract envelope of heart sounds exactly,for the purpose of the further analysis of its characteristics.Methods The way that envelope extraction of heart sounds based on key-points was given.The points of local peak and valley were calculated firstly,and then heart sound envelope was gotten by the interpolation of these points.Results Compared with the envelope extracted by Hilbert-transform and mathematical morphology,respectively,the outline of heart sounds was extracted more accurately,and its time-domain characters were acquired by this method.Conclusion The envelope of heart sound is extracted correctly by this method,which is useful for the further analysis.