Raboshchuk, G.; Nadeu, C.; Vidiella, S.; Ros, O.; Muñoz, B.; Riverola , A. Biomedical signal processing and control Vol. 39, p. 390-395 DOI: 10.1016/j.bspc.2017.07.024 Data de publicació: 2018-01-01 Article en revista
The sounds occurring in the noisy acoustical environment of a Neonatal Intensive Care Unit (NICU) are thought to affect the growth and neurodevelopment of preterm infants. Automatic sound detection in a NICU is a novel and challenging problem, and it is an essential step in the investigation of how preterm infants react to auditory stimuli of the NICU environment. In this paper, we present our work on an automatic system for detection of vocalization sounds, which are extensively present in NICUs. The proposed system reduces the presence of irrelevant sounds prior to detection. Several pre-processing techniques are compared, which are based on either spectral subtraction or non-negative matrix factorization, or a combination of both. The vocalization sounds are detected from the enhanced audio signal using either generative or discriminative classification models. An audio database acquired in a real-world NICU environment is used to assess the performance of the detection system in terms of frame-level missing and false alarm rates. The inclusion of the enhancement pre-processing step leads to up to 17.54% relative improvement over the baseline.
Balleza, J.; Pérez, E.; Vargas, J.; Kashina, S.; Huerta-Franco, M.; Torres-González, L.A.; Riu, P.J. Biomedical signal processing and control Vol. 27, p. 68-76 DOI: 10.1016/j.bspc.2016.02.001 Data de publicació: 2016-05 Article en revista
Previously, our research group obtained a set of calibration equations for being used by an electricalimpedance tomograph (EIT). All equations transform the impedance changes into a measurable volumesignal in a group of healthy males. The performance of EIT, using these equations, was acceptable forrespiration monitoring. The EIT impedance changes were obtained from a set of 16-electrodes placedaround the thoracic box at level of the sixth intercostal space. Each cycle of impedance measurementsis ordered in a matrix (IEITM). Each IEITM’s element depicts a configuration of 4-electrodes. Now, ourmain challenge is to replace the EIT’s 16-electrodes by a 4-electrodes configuration. We analyzed theimpedance changes obtained from each element and the volume determinations obtained by a pneu-motachometer (gold standard) in order to determine the optimal 4-electrodes configuration. For eachselected configuration a set of 20 calibration equations were obtained. The best results were obtained byusing two 4-electrodes configurations. Both consisting in two adjacent electrodes for current injectionand two electrodes for voltage detection.The mean R2of the 20 equations determined by the 16-electrodes information, and for the best two 4-electrode configurations were 0.943 ± 0.010, 0.848 ± 0.062 and 0.690 ± 0.122, respectively. The error (%)of volume determinations obtained by the use of 16-electrodes, and by the use of the two best 4-electrodeconfiguration with regarding to the real volume obtained by pneumotachometer were of 15 ± 6%, 16 ± 4%and 43 ± 41%, respectively. The volume differences between the one obtained by the use of 16-electrodesand that for the best 4 electrode configuration were not significant. We conclude that although thedeterminations of impedance obtained by the best 4-electrode configuration has a lower sensitivity thanthose obtained by the 16-electrodes, it is possible to measure the respiratory pattern in healthy males
This work describes an algorithm intended to detect the beat-to-beat heart rate from the ballistocardiogram (BCG) obtained from seated subjects. The algorithm is based on the continuous wavelet transform with splines, which enables the selection of an optimum scale for reducing noise and mechanical interferences. The first step of the algorithm is a learning phase in which the first four heartbeats in the BCG are detected to define initial thresholds, search windows and interval limits. The learned parameters serve to identify the next heartbeat and are readapted after each heartbeat detected to follow the heart rate and signal-amplitude changes. To evaluate the agreement between results from the algorithm and the heart rate obtained from the ECG, a Bland-Altman plot has been used to compare them for seven seated subjects. The mean error obtained was -0.03 beats/min and the 95% confidence interval (+/- 2 SD) was +/- 2.7 beats/min, which is within the accuracy limits recommended by the Association for the Advancement of Medical Instrumentation (AAMI) standard for heart rate meters. (C) 2016 Elsevier Ltd. All rights reserved.
set of calibration equations was previously obtained to transform the lung impedance changes obtained by electrical impedance tomography (EIT), using all frame's elements, into a measurable volume signal. In order to study the goodness of the use of regions of interest (ROI) for lung ventilation monitoring, we considered 6 different ROI to obtain a calibration equation for each area. Our aim was to compare the results, determined by these areas, and those obtained by using all EIT image elements. Two ROI's were defined by those pixels with an impedance change higher than 30% and 70% of the maximum change value. These areas were called P30 and P70, respectively. Two other ROI were defined by bounding two areas by mouse, resembling P30 and P70 regions, which were called M30 and M70, respectively. The remainder was defined by two elliptical areas with an eccentricity of 0.8, and 25 and 32 pixels of mayor axis (E25p and E32p, respectively). Twenty healthy males and 24 chronic obstructive pulmonary disease (COPD) patients were considered. For small region (P30 and M30) we obtained a large dispersion in volume measurement, concluding that small regions are not suitable for monitoring the tidal changes in lung volume even for healthy subjects. The results obtained by the remainder areas, and by using EIT image were similar. Even a slight improvement in data dispersion was obtained by using some ROI. These optimal results, for healthy people, were those corresponding to P70 and M70 (volume dispersion improved from 12% with the whole EIT image to 9% using ROI), and for COPD patients improves volume dispersion from 32% using the whole EIT image to 27% by using E25p. Using not so small ROI, it is possible to estimate the total lung ventilation.
Ruiz, M.; Leal, Y.; Lorencio, C.; Bondia, J.; Mujica, L.E.; Vehí, J. Biomedical signal processing and control Vol. 8, num. 6, p. 603-614 DOI: 10.1016/j.bspc.2013.05.008 Data de publicació: 2013-11-01 Article en revista
The study of motor unit action potential (MUAP) activity from electromyographic signals is important for neurological investigations aiming to understand the state of the neuromuscular system. In this context, the identification and clustering of MUAPs that exhibit common characteristics, and the assessment of which data features are most relevant for the definition of such cluster structure, are central issues. In this paper, we propose the application of an unsupervised feature relevance determination (FRD) method to the analysis of experimental MUAPs. This method is embedded in a constrained mixture of distributions model that simultaneously performs data clustering and visualization. The experimental results of the analysis of a data set consisting of MUAPs measured from the First Dorsal Interosseous, a hand muscle, indicate that the features corresponding to the hyperpolarization period in the physiological process of generating muscle fibre action potentials are consistently estimated to be the most relevant. Moreover, the MUAP cluster structure of the data is shown to be only partially attributable to inter-subject differences, with the hyperpolarization period providing the best discrimination of the data by subject.