FileName
stringlengths 17
17
| Abstract
stringlengths 163
6.01k
| Title
stringlengths 12
421
|
|---|---|---|
S1746809414000986 | An analytical study of cepstral peak prominence (CPP) is presented, intended to provide an insight into its meaning and relation with voice perturbation parameters. To carry out this analysis, a parametric approach is adopted in which voice production is modelled using the traditional source-filter model and the first cepstral peak is assumed to have Gaussian shape. It is concluded that the meaning of CPP is very similar to that of the first rahmonic and some insights are provided on its dependence with fundamental frequency and vocal tract resonances. It is further shown that CPP integrates measures of voice waveform and periodicity perturbations, be them either amplitude, frequency or noise. | Cepstral peak prominence: A comprehensive analysis |
S1746809414001001 | Monitoring the respiratory rate (RR) is important in many clinical and non-clinical situations but it is difficult in practice, for existing devices are obtrusive, bulky and expensive. The extraction of the RR from the routinely acquired electrocardiogram (ECG) has been proposed lately. Two approaches exist, one exploiting the modulation of the heart rate by the respiration, known as the respiratory sinus arrhythmia (RSA) and the other using the modulation by the respiration of the R-peak amplitudes (RPA). In this study, the weighted multi-signal oscillator based band pass filtering (W-OSC) algorithm is applied to track the common frequency in the RSA and RPA waveforms simultaneously, as an estimate of the instantaneous RR. On the public PhysioNet Fantasia data set, it is shown that the presented method is automatic, instantaneous and comparable in accuracy to the state-of-the-art. autoregressive band-pass filter breaths per minute electrocardiogram empirical mode decomposition error percentage intrinsic mode function mean absolute error oscillator based band-pass filtering R-peak amplitude respiratory rate respiratory sinus arrhythmia short time Fourier transform weighted multi-signal oscillator based band-pass filtering | Respiratory rate estimation from the ECG using an instantaneous frequency tracking algorithm |
S1746809414001104 | Originally inspired by biological neural networks, artificial neural networks (ANNs) are powerful mathematical tools that can solve complex nonlinear problems such as filtering, classification, prediction and more. This paper demonstrates the first successful implementation of ANN, specifically nonlinear autoregressive with exogenous input (NARX) networks, to estimate the hemodynamic states and neural activity from simulated and measured real blood oxygenation level dependent (BOLD) signals. Blocked and event-related BOLD data are used to test the algorithm on real experiments. The proposed method is accurate and robust even in the presence of signal noise and it does not depend on sampling interval. Moreover, the structure of the NARX networks is optimized to yield the best estimate with minimal network architecture. The results of the estimated neural activity are also discussed in terms of their potential use. | Nonlinear neural network for hemodynamic model state and input estimation using fMRI data |
S1746809414001116 | Objective Non-motor symptoms in Parkinson's disease (PD) involving cognition and emotion have been progressively receiving more attention in recent times. Electroencephalogram (EEG) signals, being an activity of central nervous system, can reflect the underlying true emotional state of a person. This paper presents a computational framework for classifying PD patients compared to healthy controls (HC) using emotional information from the brain's electrical activity. Approach Emotional EEG data were obtained from 20 PD patients and 20 healthy age-, gender- and education level-matched controls by inducing the six basic emotions of happiness, sadness, fear, anger, surprise and disgust using multimodal (audio and visual) stimuli. In addition, participants were asked to report their subjective affect. Because of the nonlinear and dynamic nature of EEG signals, we utilized higher order spectral features (specifically, bispectrum) for analysis. Two different classifiers namely K-Nearest Neighbor (KNN) and Support Vector Machine (SVM) were used to investigate the performance of the HOS based features to classify each of the six emotional states of PD patients compared to HC. Ten-fold cross-validation method was used for testing the reliability of the classifier results. Main results From the experimental results with our EEG data set, we found that (a) classification performance of bispectrum features across ALL frequency bands is better than individual frequency bands in both the groups using SVM classifier; (b) higher frequency band plays a more important role in emotion activities than lower frequency band; and (c) PD patients showed emotional impairments compared to HC, as demonstrated by a lower classification performance, particularly for negative emotions (sadness, fear, anger and disgust). Significance These results demonstrate the effectiveness of applying EEG features with machine learning techniques to classify the each emotional state difference of PD patients compared to HC, and offer a promising approach for detection of emotional impairments associated with other neurological disorders. | Detection of emotions in Parkinson's disease using higher order spectral features from brain's electrical activity |
S1746809414001128 | Parkinson's disease (PD) involves impairments of voice and speech (hypokinetic dysarthria). Dysprosody is one of the most common features of PD speech that includes alterations of rhythm and velocity of articulation. The aim of this study is the evaluation of dysprosody patterns in Parkinsonian patients during a sentence repetition task by means of a fully automated tool. Twenty PD patients (14 male and 6 female) and 19 healthy controls (9 male and 10 female) were tested. Results show significant differences between the two groups as far as the time interval between each sentence repetition (T inter), the percent of speech time with respect to sentence duration (D%) and the Net Speech Rate (NSR – defined as the number of syllables of the sentence divided by the effective speech time) are concerned. In particular, T inter is larger in PD patients while D% is higher in the control group. These results show that PD patients may exhibit longer pauses between each sentence repetition and a lower percentage of “speech time” during a whole repetition period. Thus, the decrease of D% leads to an increase of NSR. Other acoustic parameters (noise and F0 variability) did not show any significant difference. This study confirms that speech in PD patients is characterized by short rushes followed by unorthodox pauses. These results may lead to the development of a system for the automatic acoustic analysis which could significantly reduce the processing time in particular during pre-processing, that to date is a time-consuming and operator-dependent step especially in case of recordings of long duration. | Automatic identification of dysprosody in idiopathic Parkinson's disease |
S174680941400113X | We propose a methodological study for the optimization of surface EMG (sEMG)-based hand gesture classification, effective to implement a human–computer interaction device for both healthy subjects and transradial amputees. The widely commonly used unsupervised Principal Component Analysis (PCA) approach was compared to the promising supervised common spatial pattern (CSP) methodology to identify the best classification strategy and the related tuning parameters. A low density array of sEMG sensors was built to record the muscular activity of the forearm and classify five different hand gestures. Twenty healthy subjects were recruited to compute optimized parameters for (“within” analysis) and to compare between (“between” analysis) the two strategies. The system was also tested on a transradial amputee subject, in order to assess the robustness of the optimization in recognizing disabled users’ gestures. Results show that RMS-WA/ANN is the best feature vector/classifier pair for the PCA approach (accuracy 88.81±6.58%), whereas M-RMS-WA/ANN is the best pair for the CSP methodology (accuracy of 89.35±6.16%). Statistical analysis on classification results shows no significant differences between the two strategies. Moreover we found out that the optimization computed for healthy subjects was proven to be sufficiently robust to be used on the amputee subject. This motivates further investigation of the proposed methodology on a larger sample of amputees. Our results are useful to boost EMG-based hand gesture recognition and constitute a step toward the definition of an efficient EMG-controlled system for amputees. | Optimization of EMG-based hand gesture recognition: Supervised vs. unsupervised data preprocessing on healthy subjects and transradial amputees |
S1746809414001141 | Human emotional recognition using physiological signals is relatively recent. The interest to this field has been motivated by the robustness of physiological traits to avoid the artifacts created by human social masking. In this paper our attention is focus on two aims: the development of an anxiety recognition system using only one physiological signal “Blood volume pulse, BVP” and the creation of a reliable anxiety model using objective assessment. This assessment is done by using dynamics of selected features. The anxiety elicitation is based on exposure to virtual reality (EVR). The results of these investigations show that this assessment technique is effective for the construction of a reliable database and hence improves the recognition system rate. | Objective model assessment for short-term anxiety recognition from blood volume pulse signal |
S1746809414001165 | The present work proposes a system for assistance of Autistic children by analysis of eye movements. Autism is a disease characterized by abnormal eye movements and an inability to follow a pattern of object movement in different directions. Eye movement data is recorded from normal individuals over a period of five days using an Electrooculogram signal acquisition system developed in the laboratory. Hjorth Parameters are used as signal features. Eye movement directions in response to a visual stimulus for tracking an object are classified using ensemble classifiers based on bagging and adaptive boosting algorithms. Maximum classification accuracies of 83.09%, 90.27%, 80.75% and 92.27% were achieved on Hjorth Parameters as features using Bagging Ensemble classifier while tracking four different sequences. The individuals are trained by repeated tracking of the sequences such that there is an improvement in tracking over time. The system is designed to measure the tracking accuracy of following four different sequences of movement of an object in different directions as shown in a cue in a predetermined interval of time. The average tracking accuracy over ten normal subjects considering all the four sequence stimuli improves from 78.64% to 90.96% in five days which is accompanied with a decrease in staring errors from 6.36% to 1.29%. This would enable convenient detection of eye fixations/staring errors in Autistic people along with the provision of gradual improvements when the tracking sequences are not followed in 50% of the cases through consequent training. | Eye movement sequence analysis using electrooculogram to assist autistic children |
S1746809414001177 | In this paper, a detailed study on the possibility and significance of performing a parametric estimation of sample entropy (SampEn) is proposed. SampEn is a non-linear metric, meant to quantify regularity of a time series. It is widely employed on biomedical signals, especially on heart rate variability. Results relevant to approximate entropy, a related index, are also reported. An analytical expression for SampEn of an autoregressive (AR) model is derived first. Then we study the feasibility of a parametric estimation of SampEn through AR models, both on synthetic and real series. RR series of different lengths are fitted to an AR model and then expected values of SampEn (SampEn μ ) are estimated. Values of SampEn, computed from real beat-to-beat interval time series (obtained from 72 normal subjects and 29 congestive heart failure patients), with m =1 and r =0.2, are within the standard range of SampEn μ more than 83% (for series length N =75) and 28% (for N =1500) of the cases. Surrogate data have been employed to verify if departures from Gaussianity are to account for the mismatch. The work supports the finding that when numerical and parametric estimates of SampEn agree, SampEn is mainly influenced by linear properties of the series. A disagreement, on the contrary, might point those cases where SampEn is truly offering new information, not readily available with traditional temporal and spectral parameters. | Parametric estimation of sample entropy in heart rate variability analysis |
S1746809414001189 | Aiming to reduce the reconstruction time and enhance the image quality of microwave induced thermal acoustic tomography (MITAT), a new image reconstruction method named HDCS-MITAT (HDCS: hierarchical dictionary compressive sensing) is proposed. Different from the recently demonstrated CS-MITAT (CS: compressive sensing) imaging method in which only one level dictionary is applied, hierarchical dictionaries are used in the HDCS-MITAT. In this method, the dictionaries with different spatial resolutions are constructed which constitute a hierarchical structure. During the image reconstructions, first the coarsest level dictionary is utilized to roughly estimate the position of the targets in the original image domain. A reduced interested image domain can be set based on this estimation. Then the next level dictionary which has higher resolution than the above level is applied to further estimating the position of the targets and so on. Finally, the finest level dictionary is used to reconstruct the image of the targets. Compared with the CS-MITAT, this HDCS-MITAT has much less computational time and better image quality. The effectiveness of the method has been validated through some simulations and real breast tumor experiments. | Hierarchical dictionary compressive sensing (HDCS) method in microwave induced thermal acoustic tomography |
S1746809414001190 | Deceleration Capacity (DC) expresses the property of the neural control of the heart extrinsically to decelerate its rate. For the computation of DC a mathematical method has been proposed and used. Although this method was proved of significant prognostic value, it may produce meaningless negative values for DC, something in contradiction with the principle of inter beat deceleration. In this paper we propose two new methods of computation, DC sgn (DC sign) and BBDC (Beat to Beat Deceleration Capacity), which not only give positive values for DC but could also improve the original method. DC sgn modifies the filtering procedure by totally excluding from computation segments that include possible artifacts. It also uses information of four successive beats in order to detect deceleration (acceleration) segments and not only from the anchor points. BBDC bases all computations on two and not on four successive beats, detecting in this way shorter-term relationships. In order to evaluate the proposed methods, a dataset of 20 young and 20 elderly subjects, all healthy, has been used. Experimental results verify our theoretical claims and show that the proposed method can discriminate more efficiently healthy young and elderly subjects than the original method. | Deceleration Capacity of heart rate: Two new methods of computation |
S1746809414001207 | Recently, several new beamformers have been introduced for reconstruction and localization of neural sources from EEG and MEG. Although studies have compared the accuracy of beamformers for localization of strong sources in the brain, a comparison of new and conventional beamformers for time-course reconstruction of a desired source has not been previously undertaken. In this study, 8 beamformers were examined with respect to several parameters, including variations in depth, orientation, magnitude, and frequency of the simulated source to determine their (i) effectiveness at time-course reconstruction of the sources, and (ii) stability of their performances with respect to the input changes. The spatial and directional pass-bands of the beamformers were estimated via simulated and real EEG sources to determine spatial resolution. White-noise spatial maps of the beamformers were calculated to show which beamformers have a location bias. Simulated EEG data were produced by projection via forward head modelling of simulated sources onto scalp electrodes, then superimposed on real background EEG. Real EEG was recorded from a patient with essential tremor and deep brain implanted electrodes. Gain – the ratio of SNR of the reconstructed time-course to the input SNR – was the primary measure of performance of the beamformers. Overall, minimum-variance beamformers had higher Gains and superior spatial resolution to those of the minimum-norm beamformers, although their performance was more sensitive to changes in magnitude, depth, and frequency of the simulated source. White-noise spatial maps showed that several, but not all, beamformers have an undesirable location bias. | Comparison of beamformers for EEG source signal reconstruction |
S1746809414001219 | Currently, open-loop stimulation strategies are prevalent in medical bionic devices. These strategies involve setting electrical stimulation that does not change in response to neural activity. We investigate through simulation the advantages of using a closed-loop strategy that sets stimulation level based on continuous measurement of the level of neural activity. We propose a model-based controller design to control activation of retinal neurons. To deal with the lack of controllability and observability of the whole system, we use Kalman decomposition and control only the controllable and observable part. We show that the closed-loop controller performs better than the open-loop controller when perturbations are introduced into the system. We envisage that our work will give rise to more investigations of the closed-loop techniques in basic neuroscience research and in clinical applications of medical bionics. | A comparison of open-loop and closed-loop stimulation strategies to control excitation of retinal ganglion cells |
S1746809414001220 | Respiratory effort has been widely used for objective analysis of human sleep during bedtime. Several features extracted from respiratory effort signal have succeeded in automated sleep stage classification throughout the night such as variability of respiratory frequency, spectral powers in different frequency bands, respiratory regularity and self-similarity. In regard to the respiratory amplitude, it has been found that the respiratory depth is more irregular and the tidal volume is smaller during rapid-eye-movement (REM) sleep than during non-REM (NREM) sleep. However, these physiological properties have not been explicitly elaborated for sleep stage classification. By analyzing the respiratory effort amplitude, we propose a set of 12 novel features that should reflect respiratory depth and volume, respectively. They are expected to help classify sleep stages. Experiments were conducted with a data set of 48 sleepers using a linear discriminant (LD) classifier and classification performance was evaluated by overall accuracy and Cohen's Kappa coefficient of agreement. Cross validations (10-fold) show that adding the new features into the existing feature set achieved significantly improved results in classifying wake, REM sleep, light sleep and deep sleep (Kappa of 0.38 and accuracy of 63.8%) and in classifying wake, REM sleep and NREM sleep (Kappa of 0.45 and accuracy of 76.2%). In particular, the incorporation of these new features can help improve deep sleep detection to more extent (with a Kappa coefficient increasing from 0.33 to 0.43). We also revealed that calibrating the respiratory effort signals by means of body movements and performing subject-specific feature normalization can ultimately yield enhanced classification performance. | Analyzing respiratory effort amplitude for automated sleep stage classification |
S1746809414001232 | Body position changes (BPCs) are manifested as shifts in the electrical axis of the heart, which may lead to ST changes in the ECG, misclassified as ischemic events. This paper presents a novel BPC detector based on a Laplacian noise model. It is assumed that a BPC can be modelled as a step-like change in the two coefficient series that result from the Karhunen–Loève transform of the QRS complex and the ST–T segment. The generalized likelihood ratio test is explored for detection, where the statistical parameters of the Laplacian model are subject to estimation. Two databases are studied: one for assessing detection performance in healthy subjects who perform BPCs, and another for assessing the false alarm rate in ECGs recorded during percutaneous transluminal coronary angiography. The resulting probability of detection (P D ) and probability of false alarm (P F ) are 0.94 and 0.00, respectively, whereas the false alarm rate in ischemic recordings is 1event/h. The proposed detector outperforms an existing detector based on the Gaussian noise model which achieved a P D /P F of 0.90/0.01 and a false alarm rate of 2events/h. Analysis of the log-likelihood function for the Gaussian and Laplacian noise models show that latter model is more adequate. | Detection of body position changes from the ECG using a Laplacian noise model |
S1746809414001244 | The main goal of the present study was to determine the best preprocessing method for extracting the frequency-following response (FFR) in the auditory brainstem. The a posteriori Wiener filtering (APWF) method was first applied in FFR preprocessing and then compared with the standard method of conventional averaging with artifact rejection (MeanAR). Two other methods, sub-band optimal weighted averaging (SubBand) and median averaging (Median), were also investigated. FFRs were recorded from 10 normal-hearing subjects. A harmonic complex tone with a missing fundamental frequency was used as the sound stimulus. Comprehensive and quantitative indices were constructed to evaluate the quality of FFRs processed by the four methods. The indices in the time domain included the root mean square (RMS) of the residual background noise, RMS of the FFR, and autocorrelation function, and the indices in the frequency domain included the signal-to-noise ratios (SNRs) of the harmonics. The results revealed that the APWF method achieved the best performance in FFR extraction. Additionally, the effect of sweep number on FFR quality was studied. Paired t-tests indicated that APWF required far fewer sweeps compared with other methods in obtaining equivalent high-quality FFRs. In conclusion, APWF is a more suitable method for FFR preprocessing than the existing methods because of its advantages in improving SNR and experiment efficiency. | Evaluation of a posteriori Wiener filtering applied to frequency-following response extraction in the auditory brainstem |
S1746809414001256 | In this paper, a solution is proposed to predict the finger joint angle using electromyography (EMG) towards application for partial-hand amputees with functional wrist. In the experimental paradigm, the subject was instructed to continuously move one finger (middle finger for able-bodied subjects and index finger for partial-hand amputees) up to the maximum angle of flexion and extension while the wrist was conducting seven different wrist motions. A switching regime, including one linear discriminant analysis (LDA) classifier and fourteen state-space models, was proposed to continuously decode finger joint angles. LDA classifier was used to recognize which static wrist motion that the subject was conducting and choose the corresponding two state-space models for decoding joint angles of the finger with two degrees of freedom (DOFs). The average classification error rate (CER) was 6.18%, demonstrating that these seven static wrist motions along with the continuous movement of the finger could be classified. To improve the classification performance, a preprocessing method, class-wise stationary subspace analysis (cwSSA), was firstly adopted to extract the stationary components from original EMG data. Consequently, the average CER was reduced by 1.82% (p <0.05). The state-space model was adopted to estimate the finger joint angle from EMG. The average estimation performance (index R 2) of the two joint angles of the finger across seven static wrist motions achieved 0.843. This result shows that the finger's joint angles can be continuously estimated well while the wrist was conducting different static motions simultaneously. The average accuracy of seven static wrist motions with and without cwSSA and the average estimation performance of the two joint angles of the finger prove that the proposed switching regime is effective for continuous estimation of the finger joint angles under different static wrist motions from EMG. | Continuous estimation of finger joint angles under different static wrist motions from surface EMG signals |
S1746809414001268 | Operator functional state (OFS) is referred to as the ability of an operator to complete assigned tasks which may fluctuate over time. In the adaptive human–machine systems, it is required that the OFSs should be estimated in real-time in order to prevent the potential performance breakdown. To this end, an accurate OFS estimation model must be established. OFS can be reflected by operator's various physiological signals including EEG measures. In this paper, five subjects’ EEG signals were collected while working jointly with the AutoCAMS, a simulated software environment of human–machine cooperative control system. The fuzzy models are employed to estimate the OFS-related operator performance from three EEG-based input features. To derive the optimal fuzzy models, a new incremental-PID-controlled particle swarm optimization (IPID-PSO) algorithm is developed. The IPID-PSO algorithm is a combination of the standard PSO algorithm and the incremental PID control algorithm. The usefulness of the IPID-PSO algorithm is firstly validated by its application to eight benchmark function optimization problems. The superiority of IPID-PSO to the standard PSO algorithm is shown to be more significant especially for the optimization of multimodal functions with multiple local optima. Then, it is used to optimize the OFS model. The model parameters optimization accuracy and convergence rate of the IPID-PSO and standard PSO algorithm were compared when used for model-based OFS estimation. The IPID-PSO algorithm developed in this work has potential to be widely applied to other real-world optimization and model identification problems. | An incremental-PID-controlled particle swarm optimization algorithm for EEG-data-based estimation of operator functional state |
S1746809414001293 | Based on our numerical heart and chest model with stochastically modifiable action potential duration (APD) parameters, the consequences of diminished subepicardial cell-to-cell coupling were studied on beat-to-beat repolarization heterogeneity. Pathological action potential durations and transmural gradient (TG) mean values (M) were assumed in the apical segment of the five-layer heart model, while in the rest of the model the action potential parameters were kept in the normal range. APD mean values and the associated APD standard deviations (SDs) were fitted to experimental data. SD was causally related to APD mean. Repolarization heterogeneity was characterized by QRST integral maps and by the non-dipolarity index (NDI). The TG of −15 model time units (mtu) yielded an NDI of 12%. By sweeping beat-to-beat TG from −15 mtu up to +14 mtu, NDI increased from 12% up to 71%. In healthy heart M is large compared to the SD value; consequently NDI is in the stable <20% range. In arrhythmia patients TG diminishes, M and SD increase, consequently, NDI shows temporally random, increased beat-to-beat fluctuations, suitable for the characterization of repolarization heterogeneity. | Computer modelling of beat-to-beat repolarization heterogeneity in human cardiac ventricles |
S174680941400130X | Fatty liver or steatosis is a pathology characterized by fat accumulation in the liver cells. Ultrasound is the most common technique used for its evaluation, however the diagnosis is strongly dependent on the physician's expertise and system settings. These drawbacks have motivated the development of procedures for the quantitative analysis of ultrasound images to help the steatosis diagnosis. In this work, three approaches are presented and tested with human liver images. The first one addresses textural analysis of the hepatic parenchyma using five classifiers, 357 features, a feature selector, and classifiers fusion. Its performance is measured by two parameters: accuracy and area under the ROC curve. The second makes use of the hepatorenal coefficient followed by a statistical analysis to discriminate echogenicity differences between liver and kidney. The third is based on the acoustical attenuation coefficient evaluated over a line traced in the images with parallel orientation to the acoustical beam. The use of classifiers fusion has provided better results (accuracy of 0.79), when compared with the performance of the best one considered alone (0.77 for ANN). The hepatorenal coefficient proved to be a good parameter for steatosis detection with calculated sensitivity and specificity of 0.90 and 0.88, respectively. It was observed the hepatorenal coefficient is not influenced by the ultrasound machine parameters. The attenuation coefficient provided lower sensitivity and specificity values than the ones from the hepatorenal coefficient. | Detection of pathologic liver using ultrasound images |
S1746809414001311 | The early detection of abnormal heart conditions is vital to identify heart problems and avoid sudden cardiac death. The people with similar heart conditions almost have similar electrocardiogram (ECG) signals. By analyzing the ECG signals’ patterns one can predict arrhythmias. Since the conventional methods of arrhythmia detection rely on observing morphological features of the ECG signals which are tedious and very time consuming, the automatic detection of arrhythmia is more preferable. In order to automate detection of heart diseases an adequate algorithm is required which could classify the ECG signals with unknown features according to the similarities between them and the ECG signals with known features. If this classifier can find the similarities precisely, the probability of arrhythmia detection is increased and this algorithm can become a useful means in laboratories. In this article a new classification method is presented to classify ECG signals more precisely based on dynamical model of the ECG signal. In this proposed method a fuzzy classifier is constructed and its simulation results indicate that this classifier can segregate the ECGs with an accuracy of 93.34%. To further improve the performance of this classifier, genetic algorithm is applied where the accuracy in prediction is increased up to 98.67%. This proposed method increases the accuracy of the ECG classification regarding more precise arrhythmia detection. | Heart diseases prediction based on ECG signals’ classification using a genetic-fuzzy system and dynamical model of ECG signals |
S1746809414001323 | In this paper, empirical mode decomposition (EMD) is proposed as an alternative to decompose the log magnitude spectrum of the speech signal into its harmonic, envelope and noise components. The acoustic measure named harmonic-to-noise ratio (HNR) is used to summarize the degree of disturbance in the speech signal and consequently to evaluate the overall quality of the disordered voices produced by dysphonic speakers. Most approaches for HNR estimation have in common to involve the isolation of individual speech cycles or pseudo-harmonics/rhamonics in speech spectrum/cepstrum; however, this isolation cannot be carried out reliably in speech produced by severely hoarse speakers and may result in inaccurate HNR estimation. The EMD-based approach used in this study incorporates an appropriate procedure that estimates automatically the thresholds used by the clustering algorithm without knowledge of the fundamental frequency. The frequency range of the harmonic and noise components is divided into ten equally spaced intervals and the harmonic-to-noise ratios (HNRs) within each interval are used as independent variables to summarize the amount of perceived hoarseness. The proposed method is evaluated on a corpus comprising 251 normophonic and dysphonic speakers. Multiple correlation analysis carried out on HNRs from the different frequency bands shows that multi-band analysis based on empirical mode decomposition results in statistically significantly higher correlation of predicted scores with scores of perceived hoarseness over full-band analysis. Principal component analysis is carried out on the HNR measures obtained in the ten frequency bands. More than 97% of the total variance is explained by the first two principal components, PC1 and PC2. Experimental results show that the first principal component is interpretable in terms of the degree of the severity of hoarseness whereas the second principal component indicates whether the voice is high-pitched or low-pitched. It is shown that the first two principal components result in a high predictability of hoarseness scores. | Multiband vocal dysperiodicities analysis using empirical mode decomposition in the log-spectral domain |
S1746809414001335 | Detection of the nipple in mammograms is an important step in algorithms for the detection of breast cancer. However, locating the nipple position is a challenging task due to distortion and displacement of the nipple by breast diseases, improper imaging techniques, and variation of the characteristics of breast tissues with different imaging protocols or modalities. This paper presents a novel approach for the detection of the nipple in mammograms based on the converging characteristics of oriented patterns of the breast tissues towards the nipple. The oriented structures are extracted with a bank of real Gabor filters and are transformed into the Radon domain to analyze linear structures of tissue patterns to determine the nipple position. The performance of the method was evaluated with different types of images, such as scanned screen-film (from the mini-MIAS and DDSM databases), digital radiography, and computed radiography, and average errors of 7.71mm, 7.52mm, 9.23mm, and 12.10mm were achieved, respectively, with reference to the nipple location marked by an expert radiologist. The proposed method outperforms two recently developed approaches for the same application. | Detection of the nipple in mammograms with Gabor filters and the Radon transform |
S1746809414001347 | A novel and efficient signal compression algorithm aimed at finding the sparsest representation of electrocardiogram (ECG) signals is presented and analyzed. The idea behind the method relies on basis elements drawn from the initial transitory of a signal itself, and the sparsity promotion process applied to its subsequent blocks grabbed by a sliding window. The saved coefficients rescaled in a convenient range, quantized and compressed by a lossless entropy-based algorithm. Experiments on signals extracted from the MIT-BIH Arrhythmia database show that the method achieves in most of the cases very high performance. | ECG compression retaining the best natural basis k-coefficients via sparse decomposition |
S1746809414001360 | The intima–media thickness (IMT) of the common carotid artery (CCA) is being used as a reliable and early detector of atherosclerosis. Atherosclerosis may be unnoticed fo years before triggering severe illnesses such as stroke, embolisms or ischemia. Hence, the use of IMT leads to an early atherosclerosis diagnosis that can prevent more serious cardiovascular diseases. Usually, IMT is manually extracted from ultrasound images, which is a non-invasive technique, but unfortunately its measurement is prone to error. This paper addresses a fully automatic method to segment the artery layers of the CCA over ultrasound images. Unlike other methods, the segmentation is not restricted to IMT, the artery diameter can be extracted too, which can help to determine cardiovascular risk together with IMT. The proposed technique is based on a frequency-domain implementation of active contours, which are computationally faster than the original space-formulation, while providing soft final contours. Working with three different probes over a range of spatial resolutions from 0.029mm/pixel to 0.081mm/pixel, the method presents an IMT error of only 13.8±31.9μm (in mean±standard deviation) when tested on an database containing 46 images. The automatic results were compared to the average of 2 manual observations performed by 2 observers (4 observations) over each image in our database. | Frequency-domain active contours solution to evaluate intima–media thickness of the common carotid artery |
S1746809414001372 | In this work, we propose a novel phenomenological model of the EEG signal based on the dynamics of a coupled Duffing-van der Pol oscillator network. An optimization scheme is adopted to match data generated from the model with clinically obtained EEG data from subjects under resting eyes-open (EO) and eyes-closed (EC) conditions. It is shown that a coupled system of two Duffing-van der Pol oscillators with optimized parameters yields signals with characteristics that match those of the EEG in both the EO and EC cases. The results, which are reinforced using statistical analysis, show that the EEG recordings under EC and EO resting conditions are clearly distinct realizations of the same underlying model occurring due to parameter variations with qualitatively different nonlinear dynamic characteristics. In addition, the interplay between noise and nonlinearity is addressed and it is shown that, for appropriately chosen values of noise intensity in the model, very good agreement exists between the model output and the EEG in terms of the power spectrum as well as Shannon entropy. In summary, the results establish that an appropriately tuned stochastic coupled nonlinear oscillator network such as the Duffing-van der Pol system could provide a useful framework for modeling and analysis of the EEG signal. In turn, design of algorithms based on the framework has the potential to positively impact the development of novel diagnostic strategies for brain injuries and disorders. | A phenomenological model of EEG based on the dynamics of a stochastic Duffing-van der Pol oscillator network |
S1746809414001384 | Local binary pattern (LBP) is a texture descriptor that has been proven to be quite effective for various image analysis tasks in image processing. In this paper one-dimensional local binary pattern (1D-LBP) based features are used for classification of seizure and seizure-free electroencephalogram (EEG) signals. The proposed method employs a bank of Gabor filters for processing the EEG signals. The processed EEG signal is divided into smaller segments and histograms of 1D-LBPs of these segments are computed. Nearest neighbor classifier utilizes the histogram matching scores to determine whether the acquired EEG signal belongs to seizure or seizure-free category. Experimental results on publicly available database suggest that the proposed features effectively characterize local variations and are useful for classification of seizure and seizure-free EEG signals with a classification accuracy of 98.33%. This result demonstrates the superiority of our approach for classification of seizure and seizure-free EEG signals over recently proposed approaches in the literature. | Classification of seizure and seizure-free EEG signals using local binary patterns |
S1746809414001396 | Increase in intraocular pressure (IOP) is one of the causes of glaucoma which can lead to blindness if not detected and treated at an early stage. Glaucoma symptoms are not always obvious; hence patients seek treatment only when the condition progressed significantly. Early detection and treatment will decrease the chances of vision loss due to glaucoma. This paper proposes a novel automated glaucoma diagnosis method using various features extracted from Gabor transform applied on digital fundus images. In this work, we have used 510 images to classify into normal and glaucoma classes. Various features namely mean, variance, skewness, kurtosis, energy, and Shannon, Rényi, and Kapoor entropies are extracted from the Gabor transform coefficients. These extracted features are subjected to principal component analysis (PCA) to reduce the dimensionality of the features. Then these features are ranked using various ranking methods namely: Bhattacharyya space algorithm, t-test, Wilcoxon test, Receiver Operating Curve (ROC), and entropy. In this work, t-test ranking method yielded the highest performance with an average accuracy of 93.10%, sensitivity of 89.75% and specificity of 96.20% using 23 features with Support Vector Machine (SVM) classifier. Also, we have proposed a Glaucoma Risk Index (GRI) developed using principal components to classify the two classes using just one number. | Decision support system for the glaucoma using Gabor transformation |
S1746809414001402 | The P300 brain–computer interface (BCI) system relies on an oddball paradigm to elicit the P300. Besides the traditional row/column paradigm (RCP), many visual paradigms have been proposed for the P300 BCI. In our previous work, submatrix-based paradigm (SBP) was proposed and proven to be superior to the RCP in performance and user acceptability. To further improve the performance and realize an online BCI system, a dynamic SBP online BCI is proposed in this paper. The dynamic algorithm employs a threshold which dynamically limits the number of sequences. Based on analyzing the distribution characteristic of P300 in SBP, two threshold algorithms named the maximum value algorithm and the pseudo-kurtosis algorithm are proposed and compared. Online experimental results show that both dynamic algorithms can improve the performance of the SBP P300 speller, and the mean practical bit rate of the maximum value algorithm is 34.36bits/min, which is 9.04% higher than that of checkerboard paradigm (CBP) and 11.2% higher than that of Jin's n-flash adaptive system (NFA). | A dynamic submatrix-based P300 online brain–computer interface |
S1746809414001414 | Parkinson is a neurodegenerative disease, in which tremor is the main symptom. This paper investigates the use of different classification methods to identify tremors experienced by Parkinsonian patients. Some previous research has focussed tremor analysis on external body signals (e.g., electromyography, accelerometer signals, etc.). Our advantage is that we have access to sub-cortical data, which facilitates the applicability of the obtained results into real medical devices since we are dealing with brain signals directly. Local field potentials (LFP) were recorded in the subthalamic nucleus of 7 Parkinsonian patients through the implanted electrodes of a deep brain stimulation (DBS) device prior to its internalization. Measured LFP signals were preprocessed by means of splinting, down sampling, filtering, normalization and rectification. Then, feature extraction was conducted through a multi-level decomposition via a wavelet transform. Finally, artificial intelligence techniques were applied to feature selection, clustering of tremor types, and tremor detection. The key contribution of this paper is to present initial results which indicate, to a high degree of certainty, that there appear to be two distinct subgroups of patients within the group-1 of patients according to the Consensus Statement of the Movement Disorder Society on Tremor. Such results may well lead to different resultant treatments for the patients involved, depending on how their tremor has been classified. Moreover, we propose a new approach for demand driven stimulation, in which tremor detection is also based on the subtype of tremor the patient has. Applying this knowledge to the tremor detection problem, it can be concluded that the results improve when patient clustering is applied prior to detection. | Resting tremor classification and detection in Parkinson's disease patients |
S1746809414001426 | In this paper, a Hepatitis B virus model with standard incidence rate and logistic proliferation of healthy and infected cells is presented. Based on this model, we study an optimal control problem about anti-HBV infection combination therapy of Traditional Chinese Medicine and Western Medicine, the optimal strategies of taking medicine are given by simulation. Two optimal strategies with or without the impact to the infection rate by treatment are compared, simulation shows the impact to the reduction of infection may be omitted when mathematical model is used to study the anti-HBV therapy which is consistent with some references. What is more, optimal control strategy with other constant control strategies are also compared, and the simulation shows the optimal control strategy is better than constant control strategies. | Optimal control of anti-HBV treatment based on combination of Traditional Chinese Medicine and Western Medicine |
S1746809414001438 | We present a coarse-to-fine dot array marker detection algorithm which can extract dot features with high accuracy and low uncertainty. The contribution of this paper is twofold: one is a configurable dot array marker detection framework which enables real-time multi-marker tracking with compact marker size (coarse detection); the other is a closed-form sub-pixel edge localization method including the formulation and the implementation (fine localization). The marker pattern together with the dot contours is detected in a fast but coarse way for efficiency consideration, using simple thresholding and hierarchical contour analysis. If the marker pattern matches with one of predefined marker descriptors, sub-pixel edge point localization of the dot contour is performed within the detected marker region by searching the zero-crossing in the convolution of the marker image with a Laplacian-of-Gaussian (LoG) kernel. A closed-form solution is proposed to localize the “true” edge point in a 3×3 neighborhood of a candidate pixel by solving a quartic equation. The dot center is finally extracted by ellipse fitting and re-ordered according to an orientation indicator. The algorithm was evaluated against both synthetic and real image data, and also in real applications where stereo visual trackers were implemented using the proposed marker detection algorithm. Experimental results show that (1) the marker detection algorithm yielded a feature detection error of less than 0.1 pixel with real-time performance; (2) the uncertainties in both localizing static 2-D dot features and 3-D pose tracking were obviously reduced by performing the sub-pixel localization; and (3) the feasibility of the marker tracking under stereo laparoscopic views was confirmed in an in vivo animal experiment. | Coarse-to-fine dot array marker detection with accurate edge localization for stereo visual tracking |
S174680941400144X | Eye activity has larger electrical potential than the average electroencephalogram (EEG) recording, thus making it one of the major sources of artefacts. Ocular artefacts (OA) must be removed as completely as possible with little or no loss of EEG to obtain a higher quality EEG. Using independent component analysis (ICA), the EEG is separated into independent components (IC) and the contaminated component is removed, thus removing the OA. However, ICA does not separate the sources completely and some of the meaningful EEG is lost. In this paper, a new method combining ICA and wavelet neural networking (WNN) is proposed. In this method, WNN is applied to the contaminated ICs, correcting the OA and thus lowering the data lost. The method was evaluated using simulated and real datasets and the results show that the OA are successfully removed with very little data loss. | Removal of EOG artefacts by combining wavelet neural network and independent component analysis |
S1746809414001451 | Background In clinical practice, longitudinal data can be used to find trend patterns of pathema progress, such as tumour progress, along a time axis. This kind of data can be treated as time-series data. The maximum common sub-sequence is the most common method for calculating similarity of time-series data; and each point is normally treated as having the same weight. However, not all points of data within the time series should be given the same importance. According to clinical experience, the later period sub-sequence (closer to death) has a more significant effect than earlier periods in a trend analysis. Results A weighted-similarity measure based on LCSS with Constraint Window (W-LCSS-CW) Method is proposed. The results obtained from the time-series data using different weighting factors are discussed. In a study of non-small cell lung cancer using time-series data, the relative evaluation method and external evaluation method were adopted to calculate cluster effect. The results show that the proposed method, W-LCSS-CW, can improve clustering performance significantly. Clustering performance of various methods was performed using a comparison of (C index /M index ). The proposed W-LCSS-CW Method was evaluated to 1.55 which was 37.02%, 48.01%, 49.64% higher than other common methods (Euclidean, DTW, STS) respectively. Conclusions The proposed W-LCSS-CW Method is recommended for monitoring time-series data of tumour patients because the incorporated weighting factor provides more convincing cluster results for medical assist support. | Analysis of similarity measure in the longitudinal study using improved longest common subsequence method for lung cancer |
S1746809414001463 | Critically ill patients often experience stress-induced hyperglycaemia, which results in increased morbidity and mortality. Glycaemic control (GC) can be implemented in the intensive care unit (ICU) to safely manage hyperglycaemia. Two protocols SPRINT and STAR, have been implemented in the Christchurch ICU, and have been successful in treating hyperglycaemia while decreasing the risk of hypoglycaemia. This paper presents a new GC protocol that implements the probabilistic, stochastic forecasting methods of STAR, while formalizing the control methodology using model predictive control (MPC) theory to improve the ability to tune the dynamic response of the controller. This Stochastic Model Predictive (STOMP) controller predicts the response to a given insulin/nutrition intervention, and attributes weighted penalty values to several key performance metrics. The controller thus chooses an intervention at each hour that minimizes the sum of these penalties over a prediction window of 6h, which is twice as long as the 3-h window used in STAR. Clinically validated virtual trials were used to evaluate the relative performance of STOMP. Results showed STOMP was able to obtain results very similar to STAR with both protocols maintaining approximately 85% of time within 4.4–8.0mmol/L glycaemic band, and only 4–5 patients of the 149 patient STAR cohort having blood glucose (BG) <2.2mmol/L. STOMP was able to attain similar results to STAR while further increasing ease of controller tuning for different clinical requirements and reducing the number of BG measurements required by 35%. | Stochastic Model Predictive (STOMP) glycaemic control for the intensive care unit: Development and virtual trial validation |
S1746809414001475 | A major challenge for a person with diabetes is to adapt insulin dosage regimens and food intake to keep blood glucose within tolerable limits during daily life activities. The early knowledge about the effects of inputs on glycemia would provide the patients with invaluable information for appropriate on-the-spot decision making concerning the management of the disease. Against this background, in this paper we propose multi-step data-driven predictors to the purpose of predicting blood glucose multiple steps ahead in the future, supporting therefore the patients when deciding upon treatments. We formulate the predictors based on the state-space construction step in subspace identification methods for linear systems. Physiological models from the literature were used to filter the raw information on carbohydrate and insulin intakes in order to retrieve the input signals to the predictors. The clinical data of 14 type 1 diabetic patients collected in hospital during a 3-days long visit were used. Half of the data were employed for predictor development and the remaining half for validation. Mean population prediction error standard deviation on 30min, 60min, 90min, 120min ahead prediction on validation data resulted in, respectively, 19.17mg/dL, 37.99mg/dL, 50.62mg/dL and 58.06mg/dL. | Subspace-based linear multi-step predictors in type 1 diabetes mellitus |
S1746809414001487 | Segmentation is a very crucial task for the ultrasound medical images due to the presence of various imaging artifacts and noise. This paper presents a hybrid segmentation method for the ultrasound medical images that utilize both the features of the Gaussian kernel induced fuzzy C-means (GKFCM) clustering and active contour model driven by region scalable fitting (RSF) energy function. In this method, the result obtained from the GKFCM method is utilized to initialize the contour that spreads to identify the estimated regions. It also helps to estimate the several controlling parameters used in the curve evolution process. The RSF formulation that is responsible for attracting the contour toward the object boundaries removes the requirement of the re-initialization process. The performance of the proposed method is evaluated by conducting several experiments on both the synthetic and real ultrasound images. Experimental results demonstrate that the proposed method produces better results by successfully detecting the object boundaries and also ensures an improvement in segmentation accuracy compared to others. | A hybrid segmentation method based on Gaussian kernel fuzzy clustering and region based active contour model for ultrasound medical images |
S1746809414001517 | Infant crying analysis is an important tool for identifying different pathologies at a very early stage of the life of a baby. Being able to perform this task with high accuracy is therefore important and required as a medical support system to assess a baby's health. In this research we propose an automatic classification model for infant crying for early disease detection. Our model mainly consists of two phases: (a) an acoustic features acquisition from the Mel Frequency Cepstral Coefficient and the Linear Predictive Coding from signal processing and (b) the selection/creation of an optimized fuzzy model through the Genetic Selection of a Fuzzy Model (GSFM) algorithm. GSFM searches for the best model by choosing a combination of a feature selection method, a type of fuzzy processing, a learning algorithm together with its associated parameters that best fit the input data. Our approach improves the predictive accuracy on the identification of the cause of crying and clearly helps to differentiate between normal and pathological cry. Experimental results show a significant accuracy improvement when using our optimized genetic selection method for most of the cases. | Classifying infant cry patterns by the Genetic Selection of a Fuzzy Model |
S1746809414001529 | Traditional type 1 diabetes therapies are prone to show poor glucose regulation especially in the postprandial period owing to both physiological and technological limitations. Although a closed-loop controller for glucose regulation has to be tuned to minimize the postprandial excursion and avoid late hypoglycemia, the intrinsic limitations of the problem lead to a trade-off between postprandial peak and late hypoglycemia risk. This paper reveals through an intensive in-silico study with multiple controller tuning combinations that a novel safety layer for glucose controllers, the so-called SAFE loop (Revert et al., 2013), not only reduces the hypoglycemia events but also allows reducing the postprandial glucose excursion, thus breaking the implicit trade-off present in single controllers. The SAFE outer loop monitors the estimated amount of insulin on board, and modifies the control action if it is close to a unique constraint which can be adjusted with clinical criteria. A very challenging test scenario is here implemented including the rate of blood glucose appearance from intakes of mixed meals, diurnal and day-to-day time-varying metabolic changes, inherent drawbacks in sensor and actuator, and other realistic conditions. The results show a significant reduction of hypoglycemia events when SAFE is added, regardless the closed-loop glucose controller, together with a potential postprandial response improvement. | Postprandial response improvement via safety layer in closed-loop blood glucose controllers |
S1746809414001530 | Coronary artery disease (CAD) is a leading cause of death worldwide. Although coronary CT angiography (CTA) and other new technologies emerge increasingly, conventional coronary angiography (CCA) remains as the gold standard for diagnosis of CAD, and the only way to be involved in the interventional surgery. Centerline extraction of the coronary arteries is the essential information for radiologists, and is also the foundation for a computer-aided detection (CADe) system to assist them. As the data is obtained more and more, manual extraction is impractical, a fully automatic extraction method is necessary for radiologists. However, due to the projection nature, the extraction of vessels becomes extremely difficult because of non-uniform stating caused by the contrast agent distribution and overlap of the organs. Furthermore, the shape of the blood vessels is another important information needed in clinical practice, but their identification is challenging, especially at the intersectional positions. In this paper, we propose a method to extract the blood vessel contour and identify their shapes at the intersections simultaneously. Firstly, we refine Frangi's detection result to compensate the vesselness measure, ensure connectivity and eliminate artifacts as far as possible. Secondly, we study a vessel connectedness based clustering method to identify the each blood vessel. Thirdly, in order to handle the gaps and holes in enhanced vessel image, we employ a robust method based on principle curves to extract the centerlines. Finally, We evaluate the performance of our method on 60 clinical samples in angiographies. The method performs well with respect to centerline extraction, which its average accuracy is 96.247%, sensitivity is 79.981% and specificity is 97.754%. | A robust coronary artery identification and centerline extraction method in angiographies |
S1746809414001554 | Real-time classification of eye movements offers an effective mode for human–machine interaction, and many eye-based interfaces have been presented in the literature. However, such systems often require that sensors be attached around the eyes, which can be obtrusive and cause discomfort. Here, we used two electroencephalography sensors positioned over the temporal areas to perform real-time classification of eye-blink and five classes of eye movement direction. We applied a continuous wavelet transform for online detection then extracted some discriminable time-series features. Using linear classification, we obtained an average accuracy of 85.2% and sensitivity of 77.6% over all classes. The results showed that the proposed algorithm was efficient in the detection and classification of eye movements, providing high accuracy and low-latency for single trials. This work demonstrates the promise of portable eye-movement-based communication systems and the sensor positions, features extraction, and classification methods used. | Online classification algorithm for eye-movement-based communication systems using two temporal EEG sensors |
S1746809414001566 | The purpose of this study was to evaluate quantitatively the parameters of the video laryngostroboscopy (VLS) and determine correlations between the VLS parameters and acoustic vocal function parameters. Digital VLS recordings, acoustic voice assessment, calculation of dysphonia severity index (DSI) and registration of voice range profile (VRP) were performed for 206 individuals: 50 healthy and 156 patients with mass lesions of vocal folds and paralysis. 90 of lesions were unilateral, 66 – bilateral. VLS parameters were derived using objective measures made from a single image taken from the VLS recording of a sustained vowel: glottal areas, glottal widths and distances, vocal fold angles. As the result of Fisher's linear discriminant analysis, 11 VLS measurements were identified to be relevant distinguishing between normal and pathological voice groups. Correlations between the VLS parameters and results of acoustic voice analysis parameters, DSI and VRP measurements were tested using Pearson's correlation coefficient. The correlations of VLS variables and acoustic voice measurements were moderate and statistically significant. In pathological voices numerical values of VLS parameters measured reveal significant deviances from these in normal voice; therefore quantification of VLS parameters appears to be sensitive and specific distinguishing normal and pathological voices patients groups. Analysis of correlations between the quantitative measurements obtained via VLS and acoustic voice parameters provides more versatile approach into the pathophysiology of phonation and suggests the documented and measurable evidence of complex mechanisms of vocal function. | Correlation between the quantitative video laryngostroboscopic measurements and parameters of multidimensional voice assessment |
S1746809414001578 | This paper proposes a new framework to obtain quality respiratory variability signals from the raw breathing recorded in neonatal intensive care units (NICUs). It combines three consecutive blocks: an automatic rejection of artifacts, implemented by a logistic regression classifier, a two-step filtering process, and the identification of respiratory cycles, implemented by a peak detection algorithm. By means of a gold standard built from a preterm infants database, the performances of the first and third blocks have been evaluated. While the former obtains a 86% of specificity and sensitivity, the latter attains a respective 97%. The interest of our proposal in the clinical domain is illustrated by a promising application to detect promptly and non-invasively the presence of neonatal sepsis in the NICU. | Artifact rejection and cycle detection in immature breathing: Application to the early detection of neonatal sepsis |
S174680941400158X | Features greatly influence the results of speech emotion recognition, among which Mel-frequency Cepstral Coefficients (MFCC) is the most commonly used in speech emotion. However, MFCC does not consider both the relationship among neighbor coefficients of Mel filters of a frame and the relationship among coefficients of Mel filters of neighbor frames, which possibly leads to lose many useful features from spectrogram. This paper presents novel weighted spectral features based on Local Hu moments. The idea is motivated by that the energy on spectrogram would drastically vary with some emotion types such as angry and happy, while it would slightly change with other emotion types such as sadness and fear. This phenomenon would affect the local energy distribution of spectrogram in both time axis and frequency axis of spectrogram. To describe local energy distribution of spectrogram, Hu moments computed from local regions of spectrogram are used, as Hu moments can evaluate the degree how the energy is concentrated to the center of energy gravity of local region of spectrogram and can significantly vary with the speech emotion types. The conducted experiments validate the proposed features in terms of the effectiveness of the speech emotion recognition. | Weighted spectral features based on local Hu moments for speech emotion recognition |
S1746809414001591 | This paper presents a new retinal image segmentation and registration approaches. The contribution of this paper is two-fold. First, the conventional vessel-tracking methods use local sequential searching, which can be easily trapped by local intensity discontinuity or vessel rupture. The proposed method uses global graph-based decision that can segment the topological vascular tree with 1-pixel width and fully connection from retinal images. Staring from initial multi-scale ridge segmentation, the disconnected vessels are retrospectively connected and then spurious ridges are removed using a shortest path algorithm on a specially defined graph. The hypothesis testing is defined in terms of probability of pixel belong to foreground and background, which enables that the false detections could be removed. Second, the conventional point-matching methods largely depend on the branching angles of single bifurcation point. The feature correspondence across two images may not be unique due to the similar angle values. In view of this, structure-matching registration is favored. The bifurcation structure is composed of a master bifurcation point and its three connected neighboring pixels or vessel segments. The characteristic vector of each bifurcation structure consists of the normalized branching angle and length, which is fairly robust to be against translation, rotation, scaling, and even modest distortion. The experimental results are presented to demonstrate the superior performance of the proposed approach. | Retinal image registration using topological vascular tree segmentation and bifurcation structures |
S1746809414001608 | A new non-invasive method for delineation of lentigo maligna and lentigo maligna melanoma is demonstrated. The method is based on the analysis of the hyperspectral images taken in vivo before surgical excision of the lesions. For this, the characteristic features of the spectral signatures of diseased pixels and healthy pixels are extracted, which combine the intensities in a few selected wavebands with the coefficients of the wavelet frame transforms of the spectral curves. To reduce dimensionality and to reveal the internal structure of the datasets, the diffusion maps technique is applied. The averaged Nearest Neighbor and the Classification and Regression Tree (CART) classifiers are utilized as the decision units. To reduce false alarms by the CART classifier, the Aisles procedure is used. | Delineation of malignant skin tumors by hyperspectral imaging using diffusion maps dimensionality reduction |
S174680941400161X | Bipolar disorders are characterized by a mood swing, ranging from mania to depression. A system that could monitor and eventually predict these changes would be useful to improve therapy and avoid dangerous events. Speech might convey relevant information about subjects’ mood and there is a growing interest to study its changes in presence of mood disorders. In this work we present an automatic method to characterize fundamental frequency (F0) dynamics in voiced part of syllables. The method performs a segmentation of voiced sounds from running speech samples and estimates two categories of features. The first category is borrowed from Taylor's Tilt intonational model. However, the meaning of the proposed features is different from the meaning of Taylor's ones since the former are estimated from all voiced segments without performing any analysis of intonation. A second category of features takes into account the speed of change of F0. In this work, the proposed features are first estimated from an emotional speech database. Then, an analysis on speech samples acquired from eleven psychiatric patients experiencing different mood states, and eighteen healthy control subjects is introduced. Subjects had to perform a text reading task and a picture commenting task. The results of the analysis on the emotional speech database indicate that the proposed features can discriminate between high and low arousal emotions. This was verified both at single subject and group level. An intra-subject analysis was performed on bipolar patients and it highlighted significant changes of the features with different mood states, although this was not observed for all the subjects. The directions of the changes estimated for different patients experiencing the same mood swing, were not coherent and were task-dependent. Interestingly, a single-subject analysis performed on healthy controls and on bipolar patients recorded twice with the same mood label, resulted in a very small number of significant differences. In particular a very good specificity was highlighted for the Taylor-inspired features and for a subset of the second category of features, thus strengthening the significance of the results obtained with patients. Even if the number of enrolled patients is small, this work suggests that the proposed features might give a relevant contribution to the demanding research field of speech-based mood classifiers. Moreover, the results here presented indicate that a model of speech changes in bipolar patients might be subject-specific and that a richer characterization of subject status could be necessary to explain the observed variability. | Automatic analysis of speech F0 contour for the characterization of mood changes in bipolar patients |
S1746809414001621 | In this paper, a parallel-type fractional zero-phase filtering technique based on the center Grünwald–Letnikov differintegrator is proposed. We first present a left and a right Grünwald–Letnikov differintegrators, which are generalized magnitude-and-phase modulations. By using them in parallel we obtain a center Grünwald–Letnikov differintegrator, essentially a parallel-type fractional zero-phase filter. And then a center symmetrical convolution mask is constructed to implement the proposed fractional zero-phase filter. The method presented eliminates the phase distortion while offering a better compromise between signal denoising and signal information retention than conventional filtering methods. To illustrate this, the differintegrator and conventional filters were applied to electrocardiogram signals. The results indicate that the method we propose has superior performance compared with conventional denoising methods. | Parallel-type fractional zero-phase filtering for ECG signal denoising |
S1746809414001633 | Classification of electrocardiographic (ECG) signals can be deteriorated by the presence in the training set of mislabeled samples. To alleviate this issue we propose a new approach that aims at assisting the human user (cardiologist) in his/her work of labeling by removing in an automatic way the training samples with potential mislabeling problems. The proposed method is based on a genetic optimization process, in which each chromosome represents a candidate solution for validating/invalidating the training samples. Moreover, the optimization process consists of optimizing jointly two different criteria, which are the maximization of the statistical separability among classes and the minimization of the number of invalidated samples. Experimental results obtained on real ECG signals extracted from the MIT-BIH arrhythmia database confirm the effectiveness of the proposed solution. | Genetic algorithm-based method for mitigating label noise issue in ECG signal classification |
S1746809414001645 | We propose a method to use electroencephalographic (EEG) coherences as features in a brain–computer interface (BCI). The coherence provides a sense of the brain's connectivity, and it is relevant as different regions of the brain must communicate between each other for the integration of sensory information. In our case, the process of feature selection is optimized in the sense that only those statistically significant and potentially discriminative coherences at a specific frequency are used, which results in a feature vector of reduced-dimension. Next, those features are classified through an optimized linear discriminant, where the best discriminating hyperplanes are selected such that the area under the receiver operating characteristics (ROC) curve is maximized. Overall, the proposed EEG coherence selection and classification method can provide efficiency rates similar to those obtained with other methods in BCI, but with the advantage of blindly selecting and optimal combination of features out of all the possible pairwise coherences. We demonstrate the applicability of the proposed method through numerical examples using real data from motor and cognitive tasks. | An optimized feature selection and classification method for using electroencephalographic coherence in brain–computer interfaces |
S1746809414001657 | This paper proposes adaptive schemes to cope with time-related changes in muscle activities during playing video game. A myoelectric control with the core of support vector machine is applied to manipulate a car in a computer-based video game. The proposed adaptive schemes model fatigue-based changes in myoelectric signals and modify the classification criteria to keep stable performance in long-term operations. Both unsupervised and supervised methods were applied to detect time-related steady state deviations in myoelectric signal patterns. Both methods improve the performance of myoelectric control and keep it stable in long-term applications. | Adaptive myoelectric control applied to video game |
S1746809414001669 | The accurate segmentation of the optic disc (OD) offers an important cue to extract other retinal features in an automated diagnostic system, which in turn will assist ophthalmologists to track many retinopathy conditions such as glaucoma. Research contributions regarding the OD segmentation is on the rise, since the design of a robust automated system would help prevent blindness, for instance, by diagnosing glaucoma at an early stage and a condition known as ocular hypertension. Among the evaluated OD segmentation schemes, the active contour models (ACMs) have often been preferred by researchers, because ACMs are endowed with several attractive properties. To this end, we designed an OD segmentation scheme to infer how the performance of the well-known gradient vector flow (GVF) model compares with nine popular/recent ACM algorithms by supplying them with the initial OD contour derived from the circular Hough transform. The findings would hopefully equip a diagnostic system designer with an empirical support to ratify the choice of a specific model as we are bereft of such a comparative study. A dataset comprising 169 diverse retinal images was tested, and the segmentation results were assessed by a gold standard derived from the annotations of five domain experts. The segmented ODs from the GVF-based ACM coincide to a greater degree with those of the experts in 94% of the cases as predicted by the least overall Hausdorff distance value (33.49±18.21). Additionally, the decrease in the segmentation error due to the suggested ACM has been confirmed to be statistically significant in view of the p values (≤1.49e−09) from the Wilcoxon signed-rank test. The mean computational time taken by the investigated approaches has also been reported. | An empirical study on optic disc segmentation using an active contour model |
S1746809414001670 | Objectives Complexity measures have been enormously used in schizophrenia patients to estimate brain dynamics. However, the conflicting results in terms of both increased and reduced complexity values have been reported in these studies depending on the patients’ clinical status or symptom severity or medication and age status. The objective of this study is to investigate the nonlinear brain dynamics of chronic, medicated schizophrenia patients and healthy control subjects using Katz's fractal dimension (FD). Moreover, in order to determine noise effect on complexity of EEG data, a noise elimination method based on wavelet and singular spectrum analysis (SSA) were assessed. Methods Twenty-two schizophrenia patients and twenty-two age- and gender-matched control subjects underwent a resting state EEG examination with 120s. The discrete wavelet transform (DWT) was applied for EEG decomposition. Using a SSA approach, noise was removed and EEG reconstructed by inverse wavelet transform. The brain complexity of participants was investigated and compared using Katz's FD obtained from original and preprocessed EEG data. Results Lower complexity values were found in schizophrenia patients. However, this difference was only statistically significant for each channel in preprocessed, noiseless EEG data. The most significant complexity differences between patients and controls were obtained in left frontal and parietal regions of the brain. Conclusion Our findings demonstrate that the utilizing of complexity measures with preprocessing approaches on EEG data to analyze schizophrenics’ brain dynamics might be a useful and discriminative tool for diagnostic purposes. Therefore, we expect that nonlinear analysis will give us more valuable results for understanding of schizophrenics’ brain. | Investigation of the noise effect on fractal dimension of EEG in schizophrenia patients using wavelet and SSA-based approaches |
S1746809414001682 | This paper presents a novel despeckling algorithm to enhance image quality of medical ultrasound images. The proposed approach exploits coefficients in the dual-tree complex wavelet transform (DTCWT) domain to develop a new quantum-inspired thresholding function. The inter-scale correlation among the coefficients of different subbands and the intra-scale variance between the coefficient and its neighborhood in the same subband are utilized to develop a new thresholding function, which is further incorporated into a Bayesian framework to perform adaptive image despeckling. Experiments are conducted using both artificially generated and real-world medical images to demonstrate that the proposed approach outperforms the conventional image despeckling approaches. | An image despeckling approach using quantum-inspired statistics in dual-tree complex wavelet domain |
S1746809414001700 | This study investigated the impact of high-speed videoendoscopy (HSV) frame rates on the assessment of nine clinically relevant vocal-fold vibratory features. Fourteen adult patients with voice disorder and 14 adult normal controls were recorded using monochromatic rigid HSV at a rate of 16,000 frames per second (fps) and spatial resolution of 639×639 pixels. The 16,000-fps data were downsampled to 16 other rate denominations. Using paired comparisons design, nine common clinical vibratory features were visually compared between the downsampled and the original images. Three raters reported the thresholds at which: (1) a detectable difference between the two videos was first noticed, and (2) differences between the two videos would result in a change of clinical rating. Results indicated that glottal edge, mucosal wave magnitude and extent, aperiodicity, contact and loss of contact of the vocal folds were the vibratory features most sensitive to frame rate. Of these vibratory features, the glottal edge was selected for further analysis, due to its higher rating reliability, universal prevalence and consistent definition. Rates of 8000fps were found to be free from visually perceivable feature degradation, and for rates of 5333fps, degradation was minimal. For rates of 4000fps and higher, clinical assessments of glottal edge were not affected. Rates of 2000fps changed the clinical ratings in over 16% of the samples, which could lead to inaccurate functional assessment. | Experimental investigation on minimum frame rate requirements of high-speed videoendoscopy for clinical voice assessment |
S1746809414001712 | This study introduces gait asymmetry measures by comparing the ground reaction force (GRF) features of the left and right limbs. The proposed features were obtained by decomposing the GRF into components of different frequency sub-bands via the wavelet transform. The correlation coefficients between the right and left limb GRF components of the same frequency sub-band were used to characterize the degree of bilateral symmetry. The asymmetry measures were then obtained by subtracting these coefficients from one. To demonstrate the effectiveness of these asymmetry measures, the proposed measures were applied to differentiate the walking patterns of Parkinson's patients and healthy subjects. The results of the statistical analyses found that the patient group has a higher degree of gait asymmetry. By comparing these results with those obtained by conventional asymmetry measures, it was found that the proposed approach can more effectively distinguish the differences between the tested Parkinson's disease patients and the healthy control subjects. | Characterizing gait asymmetry via frequency sub-band components of the ground reaction force |
S1746809414001943 | This paper presents a new fusion scheme for the CT and MR medical images that utilizes both the features of the nonsubsampled shearlet transform (NSST) and spiking neural network. As a new image representation with the different features, the NSST is utilized to provide an effective representation of the image coefficients. Firstly, the source CT and MR images are decomposed by the NSST into several subimages. The regional energy is used to fuse the low frequency coefficients. High frequency coefficients are also fused using a pulse coupled neural network model that is used as a biologically inspired type neural network. It also retains the edges and detail information from the source images. Finally, the inverse NSST is used to produce the fused image. The performance of the proposed fusion method is evaluated by conducting several experiments on the different CT and MR medical image datasets. Experimental results demonstrate that the proposed method does not only produce better results by successfully fusing the different CT and MR images, but also ensures an improvement in the various quantitative parameters as compared to other existing methods. | Nonsubsampled shearlet based CT and MR medical image fusion using biologically inspired spiking neural network |
S1746809414001967 | In this paper, we demonstrate that spectral enhancement techniques can be configured to improve the classification accuracy of a pattern recognition-based myoelectric control system. This is based on the observation that, when the subject is at rest, the power in EMG recordings drops to levels characteristic of the noise. Two Minimum Statistics techniques, which were developed for speech processing, are compared against electromyographic (EMG) de-noising methods such as wavelets and Empirical Mode Decomposition. In the cases of simulated EMG signals contaminated with white noise and for real EMG signals with added and intrinsic noise the gesture classification accuracy was shown to increase. The mean improvement in the classification accuracy is greatest when Improved Minima-Controlled Recursive Averaging (IMCRA)-based spectral enhancement is applied, thus demonstrating the potential of spectral enhancement techniques for improving the performance of pattern recognition-based myoelectric control. | Improved pattern recognition classification accuracy for surface myoelectric signals using spectral enhancement |
S1746809414001979 | Designing transfer functions is a challenging task for medical volume data visualization, especially when an arch of the same boundary disperses seriously and adjacent arches are intersected in the intensity and gradient magnitude (IGM) transfer function space. In this paper, a novel transfer function space is proposed to better highlight and differentiate different materials in realistic volume datasets. The proposed method combines the intensity values and three-dimensional (3D) SUSAN (Smallest Univalue Segment Assimilating Nucleus) edge responses of the original data to define the intensity and SUSAN (IS) transfer function space. The results of various datasets in volume rendering show that boundary of different materials exhibits a trapezoidal shape in the proposed IS space, and boundary information is much better brought out in comparison to the IGM space. Thus the IS space provides much more intuitive clues than the IGM space in order that transfer functions can be more easily designed. Meanwhile, more details of materials of interest are visible in the rendering images. | Volume visualization based on the intensity and SUSAN transfer function spaces |
S1746809414001980 | EEG, EMG, and EOG are very informative signals recorded in polysomnography (PSG) and used for sleep staging. Their reliable acquisition at home, however, is difficult. In comparison, ECG and thoracic respiratory (R) signals are easier to record and can be useful in home sleep monitoring systems. The simultaneous utilization of Heart Rate Variability (HRV) and respiratory (R) signals seems a plausible scenario as both heart rate (HR) and respiration rate (RR) vary during different sleep states. Therefore, we explored the combined discriminative capacity (accuracy, sensitivity, and specificity) of ECG/R signals in automatic sleep staging. As baseline, we classified the wakefulness, Stage 2, SWS (slow wave sleep) and REM sleep by using a Support Vector Machine (SVM) fed with a set of features extracted from: (a) HRV (34-features), (b) HRV/ECG-Derived Respiration (45-features), and (c) combined HRV/R (45-features) signals. Approach (a) produced reasonable discriminative capacity, while approach (b) significantly improved the classification; however, the best outcomes were achieved by using approach (c). Then, we enhanced the SVM classifier with the Recursive Feature Elimination (RFE) method. The classification results were improved with 35 out of the 45 HRV/RS-EDR features. In comparison, best results were obtained by combining 27 out of the 45 features derived from HRV/R signals, in which the optimal feature set selected by the SVM-RFE method, included a combination of time domain, time-frequency, and fractal features, as well as entropies. Overall, these improvements revealed that it is possible to simplify home monitoring of sleep disorders and achieve high discriminative capacity (accuracy=89.32%, specificity=92.88%, and sensitivity=78.64%) in automatic sleep staging by the exclusive recording of cardiorespiratory signals. | Automatic sleep staging by simultaneous analysis of ECG and respiratory signals in long epochs |
S1746809414002006 | Decision tree algorithms are extensively used in machine learning field to classify biomedical signals. De-noising and feature extraction methods are also utilized to get higher classification accuracy. The goal of this study is to find an effective machine learning method for classifying ElectroMyoGram (EMG) signals by applying de-noising, feature extraction and classifier. This study presents a framework for classification of EMG signals using multiscale principal component analysis (MSPCA) for de-noising, discrete wavelet transform (DWT) for feature extraction and decision tree algorithms for classification. The presented framework automatically classifies the EMG signals as myopathic, ALS or normal, using CART, C4.5 and random forest decision tree algorithms. Results are compared by using numerous performance measures such as sensitivity, specificity, accuracy, F-measure and area under ROC curve (AUC). Combination of DWT and random forest achieved the best performance using k-fold cross-validation with 96.67% total classification accuracy. These results demonstrate that the proposed approach has the capability for the classification of EMG signals with a good accuracy. In addition, the proposed framework can be used to support clinicians for diagnosis of neuromuscular disorders. | Comparison of decision tree algorithms for EMG signal classification using DWT |
S1746809414002018 | The interpretation of physiological signals is an important subject in affective computing. In this paper, we report an experiment to collect affective galvanic skin response signals (GRS), and describe a new imbalanced fuzzy support vector machine (IBFSVM) for their classification. IBFSVM introduces denoising factors and class compensation factors, thus defining a new fuzzy membership. The effectiveness of IBFSVM is verified on various real and artificial datasets. We define an appropriate evaluation criterion (g_mean) that combines the classification accuracy of positive and negative samples, and show that IBFSVM outperforms traditional support vector machines on imbalanced datasets. By running the IBFSVM for the datasets in our experiment, we can find that the g_mean of happiness, sadness, angry and fear is 85.17%, 86.6%, 87.4%, and 81.53% respectively. So IBFSVM is an effective and feasible solution for imbalanced learning in our experiment. | Affective detection based on an imbalanced fuzzy support vector machine |
S174680941400202X | The feature extraction of event-related potentials (ERPs) is a significant prerequisite for many types of P300-BCIs. In this paper, we proposed a multi-ganglion artificial neural network based feature learning (ANNFL) method to extract a deep feature structure of single-trial multi-channel ERP signals and improve classification accuracy. Five subjects took part in the Imitating-Reading ERP experiments. We recorded the target electroencephalography (EEG) samples (elicited by target stimuli) and non-target samples (elicited by non-target stimuli) for each subjects. Then we applied ANNFL method to extract the feature vectors and classified them by using support vector machine (SVM). The ANNFL method outperforms the principal component analysis (PCA) method and conventional three-layer auto-encoder, and then leads to higher classification accuracies of five subjects’ BCI signals than using the single-channel temporal features. ANNFL is an unsupervised feature learning method, which can automatically learn feature vector from EEG data and provide more effective feature representation than PCA method and single-channel temporal feature extraction method. | Multi-ganglion ANN based feature learning with application to P300-BCI signal classification |
S1746809414002031 | A novel algorithm for the detection of fixations and smooth pursuit movements in high-speed eye-tracking data is proposed, which uses a three-stage procedure to divide the intersaccadic intervals into a sequence of fixation and smooth pursuit events. The first stage performs a preliminary segmentation while the latter two stages evaluate the characteristics of each such segment and reorganize the preliminary segments into fixations and smooth pursuit events. Five different performance measures are calculated to investigate different aspects of the algorithm's behavior. The algorithm is compared to the current state-of-the-art (I-VDT and the algorithm in [11]), as well as to annotations by two experts. The proposed algorithm performs considerably better (average Cohen's kappa 0.42) than the I-VDT algorithm (average Cohen's kappa 0.20) and the algorithm in [11] (average Cohen's kappa 0.16), when compared to the experts’ annotations. | Detection of fixations and smooth pursuit movements in high-speed eye-tracking data |
S1746809414002043 | In the inspection of treated water samples under microscope, knowing the average number of parasite (oo)cysts like Giardia and Cryptosporidium that exist in the samples is crucial as it tells whether the water is safe for consumption. Here, we introduce a new approach using a bidirectional contour tracing technique to segment and enumerate overlapping Cryptosporidium and Giardia (oo)cysts in microscopic images of treated water samples. First the image is denoised and edge detection is performed to detect the boundary of the (oo)cysts using Kirsch operator. The greyscale image is then binarized to identify the position of the (oo)cysts before it is Otsu thresholded to separate weak edge from strong edge. Then bidirectional contour tracing is implemented to isolate overlapping objects where the boundary of the (oo)cysts is traced in two different directions simultaneously. After boundary tracing, a modified ellipse fitting is executed where partial or broken ellipses can be combined to form completed ellipses that represent (oo)cysts. The proposed technique is tested on 40 FITC microscopic images containing overlapping Cryptosporidium and Giardia (oo)cysts in treated water samples. The performance of the technique is comparable to better than those of four well-known ellipse detection methods. The technique is also tested on images containing overlapping blood cells, Cryptosporidium oocysts in dirty background and rice grains, and the results are excellent. | Segmentation of overlapping Cryptosporidium and Giardia (oo)cysts using bidirectional contour tracing |
S1746809414002067 | Atherosclerosis is a major cause of coronary artery disease leading to morbidity and mortality worldwide. Currently, coronary angiography is considered to be the most accurate technique to diagnose coronary artery disease (CAD). However, this technique is an invasive and expensive procedure with risks of serious complications. Since the symptoms of CAD are not noticed until advanced stages of the disease, early and effective diagnosis of CAD is considered a pertinent measure. In this paper, a non-invasive optical signal, the finger photoplethysmogram (PPG) obtained before and after reactive hyperemia is investigated to discriminate between subjects with different CAD conditions. To this end, the PPG from both index fingers and standard 3-lead ECG of 48 patients (16 females, age 54.3±9.6 years and 32 males, age 59.9±10.6 years) scheduled for diagnostic angiography were recorded. The coronary condition of each subject was determined by three expert cardiologists (ground truth) based on these coronary angiograms. Of the 48 patients, 18 were diagnosed as having no disease (normal coronary – NC), 3 were diagnosed as having mild stenosis (MLD), 11 had single-vessel disease (SVD), 5 had two-vessel disease (2VD) and the remaining 11 were reported to have three-vessel disease (3VD). A vessel disease was determined when a significant (more than 50%) stenosis of the lumen cross-sectional area was observed. The 48 subjects were first grouped into two classes, namely high-risk: Class 1={2VD, 3VD} (N =16) and low-risk: Class 2={NC, Mild, SVD} (N =32). Using this approach, classification using a k-Nearest Neighbor classifier leads to an accuracy of 81.5%, a sensitivity of 82.0% and a specificity of 80.9%. Then all 48 subjects were regrouped slightly differently by moving the SVD subjects from the second (low-risk) to the first (high-risk) class. Therefore for the second approach high-risk: Class 1={SVD, 2VD, 3VD} (N =27), whereas low-risk: Class 2={NC, Mild} (N =21). This second approach resulted in an accuracy of 78.8%, a sensitivity of 79.3% and a specificity of 78.3%. We submit that this technique can be employed to implement an efficient triage system for scheduling coronary angiography, as it is able to identify non-invasively patients at greater risk of coronary stenosis. | Discrimination between different degrees of coronary artery disease using time-domain features of the finger photoplethysmogram in response to reactive hyperemia |
S1746809415000026 | In this paper, artifact removal from biomedical signals is addressed. We particularly focus on removing ballistocardiogram (BCG) artifact from EEG. BCG mainly appears in EEG signals during simultaneous EEG–fMRI recordings. Different from most existing artifact removal techniques, we propose a method based on dictionary learning framework. Due to strength of sparsifying dictionaries in applications such as image denoising, it is expected to succeed in BCG removal task as well. This is investigated in the proposed approach where a dictionary is learned from original EEG recording. The dictionary is designed to locally model BCG characteristics. After achieving the dictionary, BCG can be simply subtracted from the original signal and the clean EEG is obtained. Our experimental results on both synthetic and real data confirm the effectiveness of the proposed method. The results reveal the flexibility of learned dictionary for modeling the fluctuations in artifact, and removing it from original EEG signals. | EEG–fMRI: Dictionary learning for removal of ballistocardiogram artifact from EEG |
S1746809415000038 | In this paper, we present a new technique for automatic seizure detection in electroencephalogram (EEG) signals by using Hilbert marginal spectrum (HMS) analysis. As the EEG signal is highly nonlinear and nonstationary, the traditional Fourier analysis which expands signals in terms of sinusoids cannot appropriately represent the amplitude contribution from each frequency value. The HMS is derived from the empirical mode decomposition (EMD) which decomposes signal into a collection of intrinsic mode functions (IMFs). Since this decomposition is based on the local characteristic time scale of the signal, it can be well applied to nonlinear and nonstationary processes. In this work, the spectral entropies and energy features of frequency-bands of the rhythms using HMS analysis are extracted and fed into the support vector machine (SVM) for seizure detection of EEG signals. A final comparison between the results obtained with the developed technique and results adopted by Polat and coworkers using Fourier analysis with the same database is given to show the effectiveness of this technique for seizure detection. | Hilbert marginal spectrum analysis for automatic seizure detection in EEG signals |
S174680941500004X | The signal preprocessing is prerequisite for reduction of noise and for better estimation of sources from the measured field distribution of multichannel data, since different measurement channels may be contaminated by different types of artifacts and noise. Toward this, we use a combination of independent component analysis (ICA) and ensemble empirical mode decomposition (EEMD) to denoise the multichannel magnetocardiography (MCG) data. In this technique, MCG time series data is first subjected to ICA to obtain the statistically independent components (ICs) and subsequently the EEMD-interval threshold based denoising is applied to the ICs prior to the reconstruction of the signal. We compare the results obtained from EEMD-ICA with those obtained using the conventional ICA alone and also using the wavelet enhanced ICA (wICA). We illustrate the effect of these denoising techniques on the pseudo current density (PCD) maps, which aid in visualizing the source location. The results obtained from the EEMD-ICA are seen to be decidedly superior compared to those obtained by ICA alone and wICA methods. | Denoising of multichannel MCG data by the combination of EEMD and ICA and its effect on the pseudo current density maps |
S1746809415000051 | Cellular morphology and motility analysis is a key issue for abnormality identification and classification in the research of relevant biological processes. Quantitative measures are beneficial to clinicians in making their final diagnosis. This article presents a new method for measurement of live lymphocyte morphology and intracellular motion (motility) in microscopic images acquired from peripheral blood of mice post skin transplantation. Our new method explores shape, deformation and intracellular motion features of live lymphocytes. Especially, a novel way of exploiting intracellular motion information based on optical flow method is proposed. On the basis of statistical tests, optimal morphological and motility features are chosen to form a feature vector that characterizes the dynamic behavior of the lymphocytes (including shape, deformation and intercellular motion). In order to evaluate the proposed scheme, the above feature vector is used as input to a probabilistic neural network (PNN) which then classifies the dynamic behavior of lymphocytes in a set of cell image sequences into normal and abnormal categories. Comparative experiments are conducted to validate the proposed scheme, and the results revealed that joint features of shape, deformation and intracellular motion achieve the best performance in expressing the dynamic behavior of lymphocytes, compared with Fourier descriptor and Zernike moment methods. | Quantitative analysis of live lymphocytes morphology and intracellular motion in microscopic images |
S1746809415000063 | This article introduces a unique and powerful new method for spatial localization of neuronal sources that exploit the high temporal resolution of magnetoencephalography (MEG) data to locate the originating sources within the brain. A traditional frequency beamforming algorithm was adapted from its conventional application to yield information on the spatial location of simulated neuronal signals. The concept is similar to that used in signal source localization in magnetic resonance imaging (MRI) in which spatial location is determined by the frequency of oscillation of the MR signal. Whereas a traditional frequency beamformer uses the time course values of all sensors in the dataset to assign a power value for each possible frequency in the signal, it provides no information on the spatial location of those frequencies. Our approach assigns a power value to each location in the three-dimensional head volume. To compute this power value, the time courses of a subset of sensors closest to that location in space are used rather than all the time courses in the dataset. Our novel technique incorporates actual MEG sensor locations of the closest sensors at each location in space. The approach is relatively simple to implement, yields good spatial resolution, and accurately spatially locates a simulated source in low signal-to-noise environments. In this work, its performance is compared to that of the synthetic aperture magnetometry (SAM) beamformer and shown to exhibit improved spatial resolution. | Frequency-spatial beamformer for MEG source localization |
S1746809415000099 | In this study, we propose a P-wave absence (PWA) based method for atrial fibrillation (AF) identification over a short duration of electrocardiogram (ECG). The algorithm constructs a statistical model of normal sinus rhythm (SR) P-waves using a training set. Features extracted from P-waves are taken as an input to the Expectation–Maximization algorithm to create a Gaussian mixture model (GMM) of the P-wave feature space. The model is then used to identify PWA and detect AF. The algorithm performs AF identification in a single beat, and through post-processing of successive outputs using a majority voter determines the PWA over seven beats. The MIT-BIH Atrial Fibrillation Database was used to evaluate the algorithm. Classification using the majority voter showed a sensitivity of 98.09%, a specificity of 91.66%, a positive predictive value of 79.17% and an error of 6.88%. The performance of the proposed classifier is comparable to current R–R interval (RRI)-based algorithms, yet is able to detect short episodes of AF and performs rate-independent AF determination. The proposed algorithm targets atrial activity rather than ventricular activity that is targeted in RRI-based algorithms. It provides a patient specific detection of AF using a simple classifier, and can be leveraged as a tool to detect AF onsets/offsets over short AF episodes even when a patient's heart rate is controlled. | Rate-independent detection of atrial fibrillation by statistical modeling of atrial activity |
S1746809415000129 | The measurements of bio-signals are always subject to interference from noise, which would be able to affect the research results. In the present study, we introduce a technique to detect the signal quality by using ensemble empirical mode decomposition (EEMD) and Monte Carlo verification. We first decompose the original signals into several intrinsic mode functions (IMFs) and calculate the average distances between the signal IMFs and the negative (−1) slope line in Monte Carlo verification. Then, the approximate amount of white noise percent level in original signal could be obtained via corresponding to the created curve of distance and noise percent. This new proposed technique makes the approximate white noise percent level could be obtained much easier via a simple distance index through the EEMD and Monte Carlo verification methods. | Detecting signal quality by ensemble empirical mode decomposition and Monte Carlo verification |
S1746809415000142 | For the purpose of realizing an intelligent and highly accurate diagnosis system for neuro-degenerative diseases (NDD), such as amyotrophic lateral sclerosis (ALS), Parkinson's disease (PD) and Huntington's disease (HD), the present study investigated the classification capability of different gait statistical features extracted from gait rhythm signals. Nine statistical measures, including several seldom-used variability measures for these signals, were calculated for each time series. Next, after an evaluation of four popular machine learning methods, the optimal feature subset was generated with a hill-climbing feature selection method. Experiments were performed on a data set with 16 healthy control (CO) subjects, 13 ALS, 15 PD and 20 HD patients. When evaluated with the leave-one-out cross-validation (LOOCV) method, the highest accuracy rate for discriminating between groups of NDD patients and healthy control subjects was 96.83%. The best classification accuracy (100%) was obtained with two subtype binary classifiers (PD vs. CO and HD vs. CO). | Classification of gait rhythm signals between patients with neuro-degenerative diseases and normal subjects: Experiments with statistical features and different classification models |
S1746809415000154 | The aim of this study is to analyze different sources of noise in ECG recordings from stress tests in order to obtain reliable parameters and measurements for the analysis of T-wave and QRS-complex alternans (TWA & QRSA). Simple methods for eliminating common sources of noise like power-line interference, baseline drift and electromyographic noise were used. The pre-processing phase considered the detection of steep slope/spike, low amplitude signal, and flat line or missing lead artefacts. The detection of TWA and QRSA was based on Principal Component Analysis indices and wave amplitudes considering all the leads. A particular database of 106 ECG records during stress testing was considered. The signal quality analysis performed in this study has permitted to obtain reliable and noise-tolerant measurements of TWA and QRSA indices. The different diagnostic groups were used for the evaluation of the clinical significance of the alternans. Men have significantly higher values of QRSA than women. Smokers have significantly higher TWA values as compared with non-smokers. Significant negative correlation was obtained between age and both TWA and QRSA. Correlations between TWA and QRSA and the double product of arterial hypertension and the maximal heart rate during the stress test were statistically significant, positive and relatively strong. | Noise processing in exercise ECG stress test for the analysis and the clinical characterization of QRS and T wave alternans |
S1746809415000178 | The surface electromyography (sEMG) signal is a low amplitude signal that emanates from contracting muscles. It can be used directly to measure muscle activity (once noise has been removed) or it can be smoothed for some other application, e.g., orthoses or prostheses control. Here, an automatic heuristic procedure is presented which applies singular spectrum analysis (SSA) and cluster analysis to de-noise and smooth sEMG signals. SSA is a non-parametric technique that decomposes the original time series into a set of additive time series in which the noise present in the acquired signal can be easily identified. The proposed approach constitutes an alternative to the traditional smoothing procedures, such as moving average (MOVAG), root mean square (RMS), or low-pass Butterworth filtering that are used to extract the trend of the signal. To assess the quality of the method, the results of its application to a non-stationary sEMG signal are compared with those of other step-wise filtering and smoothing techniques. | An automatic SSA-based de-noising and smoothing technique for surface electromyography signals |
S174680941500018X | One of the most common signs of tiredness or fatigue is yawning. Naturally, identification of fatigued individuals would be helped if yawning is detected. Existing techniques for yawn detection are centred on measuring the mouth opening. This approach, however, may fail if the mouth is occluded by the hand, as it is frequently the case. The work presented in this paper focuses on a technique to detect yawning whilst also allowing for cases of occlusion. For measuring the mouth opening, a new technique which applies adaptive colour region is introduced. For detecting yawning whilst the mouth is occluded, local binary pattern (LBP) features are used to also identify facial distortions during yawning. In this research, the Strathclyde Facial Fatigue (SFF) database which contains genuine video footage of fatigued individuals is used for training, testing and evaluation of the system. | Yawn analysis with mouth occlusion detection |
S1746809415000191 | Registration methods have become an important tool in many medical applications. Existing methods require a good initial estimation (transformation) in order to find a global solution, i.e., if the initial estimation is far from the actual solution, incorrect solution or mismatching is very likely. In contrast, this paper presents a novel approach for globally solving the three dimensional (3D) rigid registration problem. The registration is grounded on a mathematical theory—Lipschitz optimization. It achieves a guaranteed global optimality with a rough initial estimation (e.g., even a random guess). Moreover, Munkres assignment algorithm is used to find the point correspondences. It applies the distance matrix to find an optimal correspondence. Our method is evaluated and demonstrated on MR images from porcine knees and human knees. Compared with state-of-the-art methods, the proposed technique is more robust, more accurate to perform point to point comparisons of knee cartilage thickness values for follow-up studies on the same subject. | Surface-based rigid registration using a global optimization algorithm for assessment of MRI knee cartilage thickness changes |
S1746809415000208 | An automatic voice pathology detection scheme based on the processing of speech signal is introduced that is highly reliable in detecting various vocal folds impairments. Using linear prediction (LP) analysis to accurately keep track the variations of glottal signal from speech signal is the key element of our proposed method. The linear prediction-based residual signal estimation is employed to monitor the irregularity of glottal signal with the aim of providing information about vocal folds status. The procedure is done by decomposing a voiced signal (/a/) selected from Kay Elemetrics databases using 1-D discrete wavelet decomposition in four levels, and then applying linear prediction (LP) analysis on achieved coefficients of wavelet sub-band to capture a time-frequency representation of vocal folds vibrations. Support vector machine is finally used to make a decision about the existence of any abnormality in the vocal folds of the analyzed sample. Experimental results indicate that the extracted residual signals from wavelet sub-bands provide highly reliable features especially where there are a variety of abnormal voices that are applicable for assessing voice quality and the effectiveness of the prescribed rehabilitation medicine. | Employing linear prediction residual signal of wavelet sub-bands in automatic detection of laryngeal pathology |
S174680941500021X | This review discusses the critical issues and recommended practices from the perspective of myoelectric interfaces. The major benefits and challenges of myoelectric interfaces are evaluated. The article aims to fill gaps left by previous reviews and identify avenues for future research. Recommendations are given, for example, for electrode placement, sampling rate, segmentation, and classifiers. Four groups of applications where myoelectric interfaces have been adopted are identified: assistive technology, rehabilitation technology, input devices, and silent speech interfaces. The state-of-the-art applications in each of these groups are presented. | Current state of digital signal processing in myoelectric interfaces and related applications |
S1746809415000221 | Typical fluorescence microscopy images contain large amounts of noise, which depends on the signal in a complex manner. This characteristic is substantially different from digital photography or satellite data, for which most of the existing denoising algorithms have been designed. Therefore, an efficient estimation of the noise in fluorescence micrographs and its removal pose a challenge. On the other hand, as shown previously, the use of a calibrated microscopy detector may allow computation of the signal and noise characteristics directly from the image acquisition parameters. Therefore, we propose a denoising algorithm that takes advantage of this information to obtain an estimate of the signal and the corresponding noise in the wavelet domain. This general model was constructed using actual fluorescence micrographs and utilizes intra- and inter-scale correlations of the wavelet coefficients. The signal-to-noise estimate was then applied to perform local soft thresholding in the wavelet domain. The performance of the proposed algorithm was tested using a set of images of several common subcellular structures containing various amounts of signal-dependent and signal-independent noise. The denoising performance of the new algorithm depends on the actual amount of noise and on the type of imaged structures. In every case, we demonstrated that the proposed algorithm is superior to two other locally adaptive denoising algorithms (AdaptShrink and BivarShrink) and to optimal subband adaptive soft thresholding (OraclShrink). | Application of detector precision characteristics for the denoising of biological micrographs in the wavelet domain |
S1746809415000233 | The electromyographic (EMG) signals are extensively used on feature extraction methods for movement classification purposes. High-order statistics (HOS) is being employed increasingly in myoelectric research. HOS techniques could be represented in the frequency domain (high-order spectra, e.g., bispectrum, trispectrum) or in the time domain (higher-order cumulants). More calculus is required for computing the HOS in the frequency domain. On the one hand, classical bispectrum-based features were applied to EMG signals. We propose novel third-order cumulant-based features for EMG signals. Three different classifiers are implemented for muscular-activity detection. Different analysis and evaluations were applied to both HOS-based features in order to qualify and quantify similarities. Based on these results, it is possible to conclude that cumulant-based features and bispectrum-based features had comparable behavior and allowed similar classification rates. Hence, extra calculus in order to convert time- to frequency-domain should be avoided. | On the use of high-order cumulant and bispectrum for muscular-activity detection |
S1746809415000245 | Mental stress reduces performances, on the work place and in daily life, and is one of the first causes of cognitive dysfunctions, cardiovascular disorders and depression. This study systematically reviewed existing literature investigating, in healthy subjects, the associations between acute mental stress and short term Heart Rate Variability (HRV) measures in time, frequency and non-linear domain. The goal of this study was to provide reliable information about the trends and the pivot values of HRV measures during mental stress. A systematic review and meta-analysis of the evidence was conducted, performing an exhaustive research of electronic repositories and linear researching references of papers responding to the inclusion criteria. After removing duplicates and not pertinent papers, journal papers describing well-designed studies that analyzed rigorously HRV were included if analyzed the same population of healthy subjects at rest and during mental stress. 12 papers were shortlisted, enrolling overall 758 volunteers and investigating 22 different HRV measures, 9 of which reported by at least 2 studies and therefore meta-analyzed in this review. Four measures in time and non-linear domains, associated with a normal degree of HRV variations resulted significantly depressed during stress. The power of HRV fluctuations at high frequencies was significantly depressed during stress, while the ratio between low and high frequency resulted significantly increased, suggesting a sympathetic activation and a parasympathetic withdrawal during acute mental stress. Finally, among the 15 non-linear measures extracted, only 2 were reported by at least 2 studies, therefore pooled, and only one resulted significantly depressed, suggesting a reduced chaotic behaviour during mental stress. HRV resulted significantly depressed during mental stress, showing a reduced variability and less chaotic behaviour. The pooled frequency domain measures demonstrated a significant autonomic balance shift during acute mental stress towards the sympathetic activation and the parasympathetic withdrawal. Pivot values for the pooled mean differences of HRV measures are provided. Further studies investigating HRV non-linear measures during mental stress are still required. However, the method proposed to transform and then meta-analyze the HRV measures can be applied to other fields where HRV proved to be clinically significant. | Acute mental stress assessment via short term HRV analysis in healthy adults: A systematic review with meta-analysis |
S1746809415000336 | Dual-frequency (DF) tissue harmonic imaging has been developed to take advantage of not only harmonic signal at second harmonic (2f 0) frequency but also the inter-modulation harmonic signal at fundamental (f 0) frequency for simultaneous nonlinear detection. Though phase-encoded Golay pair can improve the signal-to-noise ratio of DF harmonic signal at both f 0 and 2f 0 frequencies, conventional matched filtering cannot correctly decode the crosstalk from harmonic components at DC and third harmonic (3f 0) frequency and will lead to range side lobe artifacts in DF harmonic imaging. For orthogonal Golay pair, however, exchanging the decoding filter will output zero for the signal and keep the crosstalk the same. Therefore, the output of exchanged filtering can be subtracted from that of the original matched filtering to completely remove the spectral crosstalk. Compared to phase inversion method, the proposed orthogonal Golay decoding does not require additional transmits to cancel the unwanted DC and 3f 0 harmonic interferences and thus the achievable frame rate remains the same. Various experiments have been performed to verify the efficacy of the proposed orthogonal Golay decoding. Results from hydrophone measurements indicate that the proposed method effectively suppresses the spectral overlap between the harmonic signal and the interference. Corresponding range side lobe level (RSLL) can be suppressed by 10–20dB when the signal bandwidth is 60%. B-mode harmonic imaging also demonstrates a reduction of side lobe magnitude (SLM) by 8dB at 2f 0 frequencies. | Orthogonal Golay excitation for range side lobe elimination in dual-frequency harmonic imaging |
S1746809415000348 | The obstructive sleep apnea syndrome (OSAS) greatly affects both the health and the quality of life of children. Therefore, an early diagnosis is crucial to avoid their severe consequences. However, the standard diagnostic test (polysomnography, PSG) is time-demanding, complex, and costly. We aim at assessing a new methodology for the pediatric OSAS diagnosis to reduce these drawbacks. Airflow (AF) and oxygen saturation (SpO2) at-home recordings from 50 children were automatically processed. Information from the spectrum of AF was evaluated, as well as combined with 3% oxygen desaturation index (ODI3) through a logistic regression model. A bootstrap methodology was conducted to validate the results. OSAS significantly increased the spectral content of AF at two abnormal frequency bands below (BW1) and above (BW2) the normal respiratory range. These novel bands are consistent with the occurrence of apneic events and the posterior respiratory overexertion, respectively. The spectral information from BW1 and BW2 showed complementarity both between them and with ODI3. A logistic regression model built with 3 AF spectral features (2 from BW1 and 1 from BW2) and ODI3 achieved (mean and 95% confidence interval): 85.9% sensitivity [64.5–98.7]; 87.4% specificity [70.2–98.6]; 86.3% accuracy [74.9–95.4]; 0.947 area under the receiver-operating characteristics curve [0.826–1]; 88.4% positive predictive value [72.3–98.5]; and 85.8% negative predictive value [65.8–98.5]. The combination of the spectral information from two novel AF bands with the ODI3 from SpO2 is useful for the diagnosis of OSAS in children. | Diagnosis of pediatric obstructive sleep apnea: Preliminary findings using automatic analysis of airflow and oximetry recordings obtained at patients’ home |
S174680941500035X | When multiple acquisition systems are used to simultaneously acquire signals, synchronization issues may arise potentially causing errors in the determination of acquisition starting points and continuous clock offsets and shifts on each device. This paper introduces a processing method to efficiently synchronize these signals in the presence of white noise sources without the requirement of clock sharing or any other digital line exchange. The use of a signal source, such as white noise with a very wide frequency band, is of great interest for synchronization purposes, due to its aperiodic nature. This high bandwidth signal is simultaneously acquired by all the acquisition channels, on distinct systems, and, synchronized afterwards using cross-correlation methods. Two different correlation methods were tested; a global method, used when clock system frequencies are exactly known, and a local method, used when independent clocks evidence shifts over time that cumulatively account for long term acquisition errors in the synchronization process. In a computational simulation with known clock frequencies the results show a synchronization error of ≈1/10 of the time resolution, for both methods. For unknown clock frequencies, the global method achieved an error of 24/10 the time resolution, indicating a much poorer performance. In the experimental set-up, only the local method was tested. The best result shows a synchronization error of 4/10 of the time resolution. All the signal conditioning and acquisition parameters were chosen taking into account potential biomedical applications. | Signal (Stream) synchronization with White noise sources, in biomedical applications |
S1746809415000361 | Multichannel wireless neural signal recording systems are a prominent topic in biomedical research, but because of several limitations, such as power consumption, the device size, and enormous quantities of data, it is necessary to compress the recorded data. Compressed sensing theory can be employed to compress neural signals. However, a neural signal is usually not sparse in the time domain and contains a large number of similar non-zero points. In this article, we propose a new method for compressing not only a sparse signal but also a non-sparse signal that has identical points. First, several concepts about the identical items of the signal are introduced; thus, a method for constructing the Minimum Euclidean or Manhattan Distance Cluster-based (MDC) deterministic compressed sensing matrix is given. Moreover, the Restricted Isometry Property of the MDC matrix is supported. Third, three groups of real neural signals are used for validation. Six different random or deterministic sensing matrices under diverse reconstruction algorithms are used for the simulation. From the simulation results, it can be demonstrated that the MDC matrix can largely compress neural signals and also have a small reconstruction error. For a six-thousand-point signal, the compression rate can be up to 98%, whereas the reconstruction error is less than 0.1. In addition, from the simulation results, the MDC matrix is optimal for a signal that has an extended length. Finally, the MDC matrix can be constructed by zeros and ones; additionally, it has a simple construction structure that is highly practicable for the design of an implantable neural recording device. | Neural signal compression using a minimum Euclidean or Manhattan distance cluster-based deterministic compressed sensing matrix |
S1746809415000373 | Vectorcardiography (VCG), as an alternative to standard 12-lead electrocardiography (ECG), represents the electrical activity of the heart. Previous studies on VCG document that VCG criteria for the diagnosis of, for example, myocardial infarction (MI), ventricular hypertrophy and ischemic diseases, are superior to the corresponding 12-lead ECG criteria. Its use in clinical practice is not common because it requires the placement of additional electrodes. However, VCG leads can be derived from standard ECG by using mathematical transformations. This paper reviews the published works on transformation techniques for derivation VCG from 12-lead ECG, their historical evolution and their importance in today's clinical practice. Different kinds of criteria for evaluation accuracy of transformations are briefly described and the accuracy of individual techniques discussed. | Methods for derivation of orthogonal leads from 12-lead electrocardiogram: A review |
S1746809415000385 | The problem of recovering multi-channel EEG signals from their randomly under-sampled measurements is addressed. The objective is to reduce the energy consumed by sensing, processing and transmission in an EEG wireless body area network. Our work is based on the Blind Compressed Sensing (BCS) framework, however instead of exploiting only the sparsity of the multi-channel ensemble in a learned basis, we also make use of the ensembles’ approximate rank deficiency. Our proposed formulation requires solving new optimization problems. To solve these problems, we derive algorithms based on the Split Bregman approach. The resulting recovery results are considerably better than those of previous techniques, in terms of the quantitative and qualitative evaluations. | Energy efficient EEG sensing and transmission for wireless body area networks: A blind compressed sensing approach |
S1746809415000397 | The aim of our research was to find out if the time irreversibility as a sign of specific class of nonlinear dynamics is present even in the newborn's heart rate oscillations. Multiscale irreversibility indices (Porta's index P%, Guzik index G% and Ehlers index E) of the heart rate signals were computed in 20 healthy neonates. The presence of system nonlinearity was assessed by surrogate data analysis. The results of our analysis revealed asymmetrical nature of heart rate oscillations present in the majority of neonatal heart rate recordings. Moreover, time irreversibility index P% was able to detect shift of sympathovagal balance toward sympathetic dominance in newborns. Our findings support the concept of nonlinearity as a universal feature of the biological control system even in the early stage of the system maturation. This finding supports the application of nonlinear methods to heart rate variability analysis. | Time irreversibility of heart rate oscillations in newborns – Does it reflect system nonlinearity? |
S1746809415000403 | In this paper, motivated by the problem of multimodal retinal image registration, we introduce and improve the robust registration framework based on partial intensity invariant feature descriptor (PIIFD), then present a registration framework based on speed up robust feature (SURF) detector, PIIFD and robust point matching, called SURF–PIIFD–RPM. Existing retinal image registration algorithms are unadaptable to any case, such as complex multimodal images, poor quality, and nonvascular images. Harris-PIIFD framework usually fails in correctly aligning color retinal images with other modalities when faced large content changes. Our proposed registration framework mainly solves the problem robustly. Firstly, SURF detector is useful to extract more repeatable and scale-invariant interest points than Harris. Secondly, a single Gaussian robust point matching model is based on the kernel method of reproducing kernel Hilbert space to estimate mapping function in the presence of outliers. Most importantly, our improved registration framework performs well even when confronted a large number of outliers in the initial correspondence set. Finally, multiple experiments on our 142 multimodal retinal image pairs demonstrate that our SURF–PIIFD–RPM outperforms existing algorithms, and it is quite robust to outliers. | Robust point matching method for multimodal retinal image registration |
S1746809415000415 | Recent technological advances in wireless body sensor networks have made it possible for the development of innovative medical applications to improve health care and the quality of life. By using miniaturized wireless electroencephalography (EEG) sensors, it is possible to perform ambulatory EEG recording and real-time healthcare applications. One master consideration in using such battery-powered wireless EEG monitoring system is energy constraint at the sensor side. The traditional EEG streaming approach imposes an excessive power consumption, as it transmits the entire EEG signals wirelessly. Therefore, innovative solutions to alleviate the total power consumption at the receiver are highly desired. This work introduces the use of the discrete wavelet transform and compressive sensing algorithms for scalable EEG data compression in wireless sensors in order to address the power and distortion constraints. Encoding and transmission power models of both systems are presented which enable analysis of power and performance costs. We then present a theoretical analysis of the obtained distortion caused by source encoding and channel errors. Based on this analysis, we develop an optimization scheme that minimizes the total distortion for different channel conditions and encoder settings. Using the developed framework, the encoder can adaptively tune the encoding parameters to match the energy constraint without performance degradation. | Scalable real-time energy-efficient EEG compression scheme for wireless body area sensor network |
S1746809415000427 | The present study was designed to achieve a comprehensive analysis of gender-related differences in the myoelectric activity of lower limb muscles during normal walking at self-selected speed and cadence, in terms of muscle activation patterns and occurrence frequencies. To this aim, statistical gait analysis (SGA) of surface EMG signal from tibialis anterior (TA), gastrocnemius lateralis (GL), rectus femoris (RF), biceps femoris (BF) and vastus lateralis (VL) was performed in 11 female (F-group) and 11 male (M-group) age-matched healthy young adults. SGA is a recent methodology performing a statistical characterization of gait, by averaging spatio-temporal and sEMG-based parameters over numerous strides. Findings showed that males and females walk at the same comfortable speed, despite the significantly lower height and higher cadence detected in females. No significant differences in muscle onset/offset were detected between groups. The analysis of occurrence frequencies of muscle activity showed no significant differences in BF and RF, between groups. Conversely, in F-group, compared with M-group, GL, TA and VL showed a significantly higher occurrence frequency in the modalities with a high number of activations, and a significantly lower occurrence frequency in the modalities with a low number of activations. These findings indicate a propensity of females for a more complex recruitment of TA, GL and VL during walking, compared to males. The observed differences recommend the suitability of developing electromyographic databases, separated for males and females. | Gender differences in the myoelectric activity of lower limb muscles in young healthy subjects during walking |
S1746809415000439 | An analysis of newborn cry signals, either for the early diagnosis of neonatal health problems or to determine the category of a cry (e.g., pain, discomfort, birth cry, and fear), requires a primary and preliminary preprocessing step to quantify the important expiratory and inspiratory parts of the audio recordings of newborn cries. Data typically contain clean cries interspersed with sections of other sounds (generally, the sounds of speech, noise, or medical equipment) or silence. The purpose of signal segmentation is to differentiate the important acoustic parts of the cry recordings from the unimportant acoustic activities that compose the audio signals. This paper reports on our research to establish an automatic segmentation system for newborn cry recordings based on Hidden Markov Models using the HTK (Hidden Markov Model Toolkit). The system presented in this report is able to detect the two basic constituents of a cry, which are the audible expiratory and inspiratory parts, using a two-stage recognition architecture. The system is trained and tested on a real database collected from normal and pathological newborns. The experimental results indicate that the system yields accuracies of up to 83.79%. | Automatic detection of the expiratory and inspiratory phases in newborn cry signals |
S1746809415000440 | The growing interest in wearable computing during daily life has lead to many studies on unconstrained biological signal measurements. The photoplethysmography (PPG), as an extremely useful wearable sensing medical diagnostic tool, adequately creates a health care monitoring device since it can be easily measured in our bodies. In this paper, we study the decomposition of photoplethysmography signal based on a finite Gaussian basis. When we employ a set of n (n <8) Gaussian basis to approximate the original PPG signal, we can use a feature vector only including 3n parameters to represent the original PPG signal, with almost no losses in geometrical shape. In contrast with a thousand samples in time domain, the proposed method can save a lot of resources in processing, transmitting and storing PPG signal. Besides that, we studied the application of our decomposition method for the extraction of respiratory and heart information from PPG signal. Determination of baseline heart rate and respiratory rate were easily identified in the experiments of exercise condition. The results indicate the accurate determination of heart rate and respiratory rate from PPG signal. We believe that method could soon be easily incorporated into current Body Area Network applications. | A new signal decomposition to estimate breathing rate and heart rate from photoplethysmography signal |
S1746809415000452 | In this paper the fetal electrocardiogram (fECG) enhancement in abdominal recordings acquired by electrodes placed on the maternal belly is considered, a new algorithm being proposed to remove the power line inference (PLI). Typically, PLI affects the evaluation of physiological signals, e.g., (fECG), for diagnostic purposes, therefore different hardware and software approaches to reduce/remove it from biomedical measurements have been proposed, some considering even the case of very low energy physiological signals. The remaining PLI still impairs the analysis of the signal of interest in some specific cases like fetal monitoring, where the morphology of fECG is essential. This paper proposes an adaptive filter based on Hilbert Huang Transform (HHT) to remove both the PLI fundamental frequency and its harmonics from the abdominal recordings, allowing the further fECG processing. The proposed algorithm, the Internal Powerline Reference Adaptive Canceler (IPRAC), shows very good performance in PLI cancellation without affecting the fECG morphology. The IPRAC performance is evaluated on both real and simulated signals including also the worst case scenario when the PLI signal does not have an absolutely constant fundamental frequency. It outperforms two recently investigated algorithms as proved by the evaluation of four different quantitative performance indexes analyzed in this study. | Fetal ECG enhancement: Adaptive power line interference cancellation based on Hilbert Huang Transform |
S1746809415000464 | In recent years, the recommendations of the Association for the Advancement of Medical Instrumentation (AAMI) for class labeling and results presentation are closely followed as a possible solution for standardization. Regardless of the class normalization, this standard basically recommends for performance evaluation to adopt inter-patient scenarios, which renders the classification task very challenging due to the strong variability of ECG signals. To deal with this issue, we propose in this paper a novel interactive ensemble learning approach based on the extreme learning machine (ELM) classifier and the induced ordered weighted averaging (IOWA) operators. While ELM is adopted for ensemble generation the IOWA operators are used for aggregating the obtained predictions in a nonlinear way. During the iterative learning process, the approach allows the expert to label the most relevant and uncertain ECG heart beats in the data under analysis and then adds them to the original training set for retraining. The experimental results obtained on the widely used MIT-BIH arrhythmia database show that the proposed approach significantly outperforms state-of-the-art methods after labeling on average 100 ECG beats per record. In addition, the results obtained on four other ECG databases starting with the same initial training set from MIT-BIH confirm its promising generalization capability. | Classification of AAMI heartbeat classes with an interactive ELM ensemble learning approach |
S1746809415000476 | The nuchal translucency thickness is an important parameter for the diagnosis of fetal abnormalities during 11–14 weeks. Currently in clinical practice, it first requires manual scanning operations to determine the fetal standard sagittal-view plane and then the measurements can be performed in the corresponding plane images. Besides the difficulty of such standard plane detection, this also leads to the time-consuming and detection-variability problems. In the paper, a hierarchical model is proposed to automatically detect the standard sagittal-view plane based on 3D ultrasound data. In the model, Hessian-matrix based filtering is first applied for obtaining the plate-structure distribution in the data. Then the sphere distribution is calculated by convolving sphere detectors with the ultrasound data. Based on the two prior distributions, the sampling-based Hough transform is further performed for the plane detection. The performance of the proposed model is verified by the experimental results on 3D synthetic data and 241 clinical 3D ultrasound data in 11–14 weeks. | A hierarchical model for automated standard sagittal-view detection from 3D ultrasound data in 11–14 weeks |
S1746809415000488 | Functional electrical stimulation (FES) applied to tibialis anterior (TA) has been proven as a useful intervention for gait rehabilitation. Because stimulation parameters in FES-assisted walking are associated with TA activation, the objective of this study was to characterize TA electromyography (EMG) among four terrains, including normal walking, obstacle crossing, and stairs ascending and descending. Six healthy subjects participated in the experiment and TA EMG, lower limb kinematic and kinetic data were recorded simultaneously when subjects were instructed to walk on the four terrains. The results showed that there was a significant difference in the time interval from the heel-off event to the onset timing of TA activation when stairs descending was compared with normal walking, obstacle crossing and stairs ascending (P <0.05). Significant differences in EMG duration between each pair of terrains were also found (P <0.05). The normalized peak EMG amplitude of stairs ascending was significantly larger than normal walking and stairs descending (P <0.05). When terrains changed, TA activation was altered by the central nervous system for a stable landing and a forward body movement. These results may provide evidence for terrain-adaptive parameters in FES-assisted walking to correct dropfoot. | Effect of different terrains on onset timing, duration and amplitude of tibialis anterior activation |
S174680941500049X | In this paper, we propose a novel approach for the compression of multichannel electroencephalograph (EEG) signals. The method assumes that EEG signals are the linear mixture of several independent components (ICs). To retain the ICs, the proposed scheme first applies an independent component analysis (ICA) with a preprocessing step of principal component analysis (PCA) to EEG signals. Then the compression scheme is composed of two parts: the ICs compression part and the residue compression part. Each IC is arranged in the form of matrix and then compressed with the algorithm of set partitioning in hierarchical trees (SPIHT). The residue signals are compressed in the same way as ICs, but with a higher compression ratio (CR). The appropriate combination of compression ratios of the ICs and the residue is explored to achieve desired performance. The compression scheme is tested with eight datasets sampled at two different frequencies. The experimental results demonstrate the high compression performance of the proposed approach and its potential usage in the EEG related telemedicine applications. | Multichannel EEG compression based on ICA and SPIHT |
Subsets and Splits