Sleep Apnea-Specific Hypoxic Burden, Symptom Subtypes, and Risk of Cardiovascular Events and All-Cause Mortality.

Rationale: Data from population-based cohorts suggest that symptom subtypes and obstructive sleep apnea (OSA)-specific hypoxic burden (HB) could help to better identify patients with OSA at high cardiovascular (CV) risk. Objectives: We aimed to evaluate whether those new markers are associated with the risk of major adverse CV events (MACE) in clinical setting. Methods: Data from the Pays de la Loire cohort were linked to health administrative data to identify the occurrence of MACE (a composite outcome including all-cause mortality, acute myocardial infarction, stroke, and unplanned coronary revascularization) in patients with newly diagnosed OSA and no overt CV disease. Latent class analysis was used to identify subtypes based on eight clinically relevant variables. HB was defined as the total area under the respiratory event-related desaturation curve. Cox proportional hazards models were used to evaluate the association of symptom subtypes and HB with MACE. Measurements and Main Results: Four symptom subtypes were identified (minimally symptomatic [22.0%], disturbed sleep [17.5%], excessively sleepy [49.8%], and moderately sleepy [10.6%]). After a median follow-up of 78 months (interquartile range, 52-109), 592 (11.05%) of 5,358 patients experienced MACE. In a fully adjusted model, HB and overall nocturnal hypoxemia assessed by sleep time with oxygen saturation <90% were the only predictors of MACE (hazard ratio, 1.21; 95% confidence interval, 1.07-1.38; and hazard ratio, 1.34; 95% confidence interval, 1.16-1.55, respectively). The association appeared stronger toward younger patients and women. Conclusion: In clinical setting, patients with OSA who demonstrate elevated OSA-specific HB are at higher risk of a CV event and all-cause mortality. Symptom subtypes were not associated with MACE after adjustment for confounders.

Palindromic Vectors, Symmetropy and Symmentropy as Symmetry Descriptors of Binary Data.

Today, the palindromic analysis of biological sequences, based exclusively on the study of "mirror" symmetry properties, is almost unavoidable. However, other types of symmetry, such as those present in friezes, could allow us to analyze binary sequences from another point of view. New tools, such as symmetropy and symmentropy, based on new types of palindromes allow us to discriminate binarized 1/f noise sequences better than Lempel-Ziv complexity. These new palindromes with new types of symmetry also allow for better discrimination of binarized DNA sequences. A relative error of 6% of symmetropy is obtained from the HUMHBB and YEAST1 DNA sequences. A factor of 4 between the slopes obtained from the linear fits of the local symmentropies for the two DNA sequences shows the discriminative capacity of the local symmentropy. Moreover, it is highlighted that a certain number of these new palindromes of sizes greater than 30 bits are more discriminating than those of smaller sizes assimilated to those from an independent and identically distributed random variable.

Centered and Averaged Fuzzy Entropy to Improve Fuzzy Entropy Precision.

Several entropy measures are now widely used to analyze real-world time series. Among them, we can cite approximate entropy, sample entropy and fuzzy entropy (FuzzyEn), the latter one being probably the most efficient among the three. However, FuzzyEn precision depends on the number of samples in the data under study. The longer the signal, the better it is. Nevertheless, long signals are often difficult to obtain in real applications. This is why we herein propose a new FuzzyEn that presents better precision than the standard FuzzyEn. This is performed by increasing the number of samples used in the computation of the entropy measure, without changing the length of the time series. Thus, for the comparisons of the patterns, the mean value is no longer a constraint. Moreover, translated patterns are not the only ones considered: reflected, inversed, and glide-reflected patterns are also taken into account. The new measure (so-called centered and averaged FuzzyEn) is applied to synthetic and biomedical signals. The results show that the centered and averaged FuzzyEn leads to more precise results than the standard FuzzyEn: the relative percentile range is reduced compared to the standard sample entropy and fuzzy entropy measures. The centered and averaged FuzzyEn could now be used in other applications to compare its performances to those of other already-existing entropy measures.

A New Sleep Staging System for Type III Sleep Studies Equipped With a Tracheal Sound Sensor.

Type III sleep studies record cardio-respiratory channels only. Compared with polysomnography, which also records electrophysiological channels, they present many advantages: they are less expensive, less time-consuming, and more likely to be performed at home. However, their accuracy is limited by missing sleep information. That is why many studies present specific cardio-respiratory parameters to assess the causal effects of sleep stages upon cardiac or respiratory activities. For this paper, we gathered many parameters proposed in literature, leading to 1,111 features. The pulse oximeter, the PneaVoX sensor (recording tracheal sounds), respiratory inductance plethysmography belts, the nasal cannula and the actimeter provided the 112 worthiest ones for automatic sleep scoring. Then, a 3-step model was implemented: classification with a multi-layer perceptron, sleep transition rules corrections (from the AASM guidelines), and sequence corrections using a Viterbi hidden Markov model. The whole process was trained and tested using 300 and 100 independent recordings provided from patients suspected of having sleep breathing disorders. Results indicated that the system achieves substantial agreement with manual scoring for classifications into 2 stages (wake vs. sleep: mean Cohen's Kappa κ of 0.63 and accuracy rate Acc of 87.8%) and 3 stages (wake vs. R stage vs. NREM stage: mean κ of 0.60 and Acc of 78.5%). It indicates that the method could provide information to help specialists while diagnosing sleep. The presented model had promising results and may enhance clinical diagnosis.

Prediction of protein structural class based on symmetrical recurrence quantification analysis.

Protein structural class prediction for low similarity sequences is a significant challenge and one of the deeply explored subjects. This plays an important role in drug design, folding recognition of protein, functional analysis and several other biology applications. In this paper, we worked with two benchmark databases existing in the literature (1) 25PDB and (2) 1189 to apply our proposed method for predicting protein structural class. Initially, we transformed protein sequences into DNA sequences and then into binary sequences. Furthermore, we applied symmetrical recurrence quantification analysis (the new approach), where we got 8 features from each symmetry plot computation. Moreover, the machine learning algorithms such as Linear Discriminant Analysis (LDA), Random Forest (RF) and Support Vector Machine (SVM) are used. In addition, comparison was made to find the best classifier for protein structural class prediction. Results show that symmetrical recurrence quantification as feature extraction method with RF classifier outperformed existing methods with an overall accuracy of 100% without overfitting.

Cardiovascular risk and mortality prediction in patients suspected of sleep apnea: a model based on an artificial intelligence system.

Objective. Cardiovascular disease (CVD) is one of the leading causes of death worldwide. There are many CVD risk estimators but very few take into account sleep features. Moreover, they are rarely tested on patients investigated for obstructive sleep apnea (OSA). However, numerous studies have demonstrated that OSA index or sleep features are associated with CVD and mortality. The aim of this study is to propose a new simple CVD and mortality risk estimator for use in routine sleep testing.Approach. Data from a large multicenter cohort of CVD-free patients investigated for OSA were linked to the French Health System to identify new-onset CVD. Clinical features were collected and sleep features were extracted from sleep recordings. A machine-learning model based on trees, AdaBoost, was applied to estimate the CVD and mortality risk score.Main results. After a median [inter-quartile range] follow-up of 6.0 [3.5-8.5] years, 685 of 5234 patients had received a diagnosis of CVD or had died. Following a selection of features, from the original 30 features, 9 were selected, including five clinical and four sleep oximetry features. The final model included age, gender, hypertension, diabetes, systolic blood pressure, oxygen saturation and pulse rate variability (PRV) features. An area under the receiver operating characteristic curve (AUC) of 0.78 was reached.Significance. AdaBoost, an interpretable machine-learning model, was applied to predict 6 year CVD and mortality in patients investigated for clinical suspicion of OSA. A mixed set of simple clinical features, nocturnal hypoxemia and PRV features derived from single channel pulse oximetry were used.

Association of Nocturnal Hypoxemia and Pulse Rate Variability with Incident Atrial Fibrillation in Patients Investigated for Obstructive Sleep Apnea.

Rationale: Nocturnal hypoxemia and sympathetic/parasympathetic imbalance might contribute to the occurrence or atrial fibrillation (AF) in patients with obstructive sleep apnea (OSA). During sleep recordings, pulse rate variability (PRV) derived from oximetry might provide an accurate estimation of heart rate variability, which reflects the autonomic cardiovascular control. Objectives: We aimed to evaluate whether indices of oxygen desaturation and PRV derived from nocturnal oximetry were associated with AF incidence in patients investigated for OSA. Methods: Data from a large multicenter cohort of AF-free patients investigated for OSA between May 15, 2007, and December 31, 2017, were linked to health administrative data to identify hospitalized and nonhospitalized patients with new-onset AF. Cox proportional hazards models were used to evaluate the association between AF incidence and oximetry-derived indices automatically generated from sleep recordings. Results: After a median (interquartile range) follow-up of 5.34 (3.3-8.0) years, 181 of 7,205 patients developed AF (130 were hospitalized for AF). After adjusting for confounders, including anthropomorphic data, alcohol intake, cardiac, metabolic and respiratory diseases, ß blocker/calcium channel blocker medications, type of sleep study, study site, and positive airway pressure adherence, AF risk was associated with increasing nocturnal hypoxemia (P trend = 0.004 for quartiles of percentage of recording time with oxygen saturation <90%) and PRV (P trend < 0.0001 for quartiles of root mean square of the successive normal-normal beat interval differences), and decreasing sympathetic/parasympathetic tone (P trend = 0.0006 for quartiles of low-frequency power/high-frequency power ratio). The highest risk of AF was observed in patients with the highest quartiles of both the percentage of recording time with oxygen saturation <90% and the root mean square of the successive normal-normal beat interval differences compared with those with neither of these conditions (adjusted hazard ratio, 3.61; 95% confidence interval, 2.10-6.22). Similar associations were observed when the analyses were restricted to hospitalized AF. Conclusions: In patients investigated for OSA, nocturnal hypoxemia and PRV indices derived from single-channel pulse oximetry were independent predictors of AF incidence. Patients with both marked nocturnal hypoxemia and high PRV were at higher risk of AF. Oximetry may be used to identify patients with OSA at greatest risk of developing AF.

Adaptive AR and neurofuzzy approaches: access to cerebral particle signatures.

In recent years, a relationship has been suggested between the occurrence of cerebral embolism and stroke. Ultrasound has therefore become essential in the detection of emboli when monitoring cerebral vascular disorders and forms part of ultrasound brain-imaging techniques. Such detection is based on investigating the middle cerebral artery using a TransCranial Doppler (TCD) system, and analyzing the Doppler signal of the embolism. Most of the emboli detected in practical experiments are large emboli because their signatures are easy to recognize in the TCD signal. However, detection of small emboli remains a challenge. Various approaches have been proposed to solve the problem, ranging from the exclusive use of expert human knowledge to automated collection of signal parameters. Many studies have recently been performed using time-frequency distributions and classical parameter modeling for automatic detection of emboli. It has been shown that autoregressive (AR) modeling associated with an abrupt change detection technique is one of the best methods for detection of microemboli. One alternative to this is a technique based on taking expert knowledge into account. This paper aims to unite these two approaches using AR modeling and expert knowledge through a neurofuzzy approach. The originality of this approach lies in combining these two techniques and then proposing a parameter referred to as score ranging from 0 to 1. Unlike classical techniques, this score is not only a measure of confidence of detection but also a tool enabling the final detection of the presence or absence of microemboli to be performed by the practitioner. Finally, this paper provides performance evaluation and comparison with an automated technique, i.e., AR modeling used in vitro.

Optimization of contrast resolution by genetic algorithm in ultrasound tissue harmonic imaging.

The development of ultrasound imaging techniques such as pulse inversion has improved tissue harmonic imaging. Nevertheless, no recommendation has been made to date for the design of the waveform transmitted through the medium being explored. Our aim was therefore to find automatically the optimal "imaging" wave which maximized the contrast resolution without a priori information. To overcome assumption regarding the waveform, a genetic algorithm investigated the medium thanks to the transmission of stochastic "explorer" waves. Moreover, these stochastic signals could be constrained by the type of generator available (bipolar or arbitrary). To implement it, we changed the current pulse inversion imaging system by including feedback. Thus the method optimized the contrast resolution by adaptively selecting the samples of the excitation. In simulation, we benchmarked the contrast effectiveness of the best found transmitted stochastic commands and the usual fixed-frequency command. The optimization method converged quickly after around 300 iterations in the same optimal area. These results were confirmed experimentally. In the experimental case, the contrast resolution measured on a radiofrequency line could be improved by 6% with a bipolar generator and it could still increase by 15% with an arbitrary waveform generator.

Detection of Doppler Microembolic Signals Using High Order Statistics.

Robust detection of the smallest circulating cerebral microemboli is an efficient way of preventing strokes, which is second cause of mortality worldwide. Transcranial Doppler ultrasound is widely considered the most convenient system for the detection of microemboli. The most common standard detection is achieved through the Doppler energy signal and depends on an empirically set constant threshold. On the other hand, in the past few years, higher order statistics have been an extensive field of research as they represent descriptive statistics that can be used to detect signal outliers. In this study, we propose new types of microembolic detectors based on the windowed calculation of the third moment skewness and fourth moment kurtosis of the energy signal. During energy embolus-free periods the distribution of the energy is not altered and the skewness and kurtosis signals do not exhibit any peak values. In the presence of emboli, the energy distribution is distorted and the skewness and kurtosis signals exhibit peaks, corresponding to the latter emboli. Applied on real signals, the detection of microemboli through the skewness and kurtosis signals outperformed the detection through standard methods. The sensitivities and specificities reached 78% and 91% and 80% and 90% for the skewness and kurtosis detectors, respectively.

Reducing sojourn points from recurrence plots to improve transition detection: Application to fetal heart rate transitions.

The analysis of biomedical signals demonstrating complexity through recurrence plots is challenging. Quantification of recurrences is often biased by sojourn points that hide dynamic transitions. To overcome this problem, time series have previously been embedded at high dimensions. However, no one has quantified the elimination of sojourn points and rate of detection, nor the enhancement of transition detection has been investigated. This paper reports our on-going efforts to improve the detection of dynamic transitions from logistic maps and fetal hearts by reducing sojourn points. Three signal-based recurrence plots were developed, i.e. embedded with specific settings, derivative-based and m-time pattern. Determinism, cross-determinism and percentage of reduced sojourn points were computed to detect transitions. For logistic maps, an increase of 50% and 34.3% in sensitivity of detection over alternatives was achieved by m-time pattern and embedded recurrence plots with specific settings, respectively, and with a 100% specificity. For fetal heart rates, embedded recurrence plots with specific settings provided the best performance, followed by derivative-based recurrence plot, then unembedded recurrence plot using the determinism parameter. The relative errors between healthy and distressed fetuses were 153%, 95% and 91%. More than 50% of sojourn points were eliminated, allowing better detection of heart transitions triggered by gaseous exchange factors. This could be significant in improving the diagnosis of fetal state.

Multidimensional Ultrasound Doppler Signal Analysis for Fetal Activity Monitoring.

Fetal activity parameters such as movements, heart rate and the related parameters are essential indicators of fetal wellbeing, and no device provides simultaneous access to and sufficient estimation of all of these parameters to evaluate fetal health. This work was aimed at collecting these parameters to automatically separate healthy from compromised fetuses. To achieve this goal, we first developed a multi-sensor-multi-gate Doppler system. Then we recorded multidimensional Doppler signals and estimated the fetal activity parameters via dedicated signal processing techniques. Finally, we combined these parameters into four sets of parameters (or four hyper-parameters) to determine the set of parameters that is able to separate healthy from other fetuses. To validate our system, a data set consisting of two groups of fetal signals (normal and compromised) was established and provided by physicians. From the estimated parameters, an instantaneous Manning-like score, referred to as the ultrasonic score, was calculated and was used together with movements, heart rate and the associated parameters in a classification process employing the support vector machine method. We investigated the influence of the sets of parameters and evaluated the performance of the support vector machine using the computation of sensibility, specificity, percentage of support vectors and total classification error. The sensitivity of the four sets ranged from 79% to 100%. Specificity was 100% for all sets. The total classification error ranged from 0% to 20%. The percentage of support vectors ranged from 33% to 49%. Overall, the best results were obtained with the set of parameters consisting of fetal movement, short-term variability, long-term variability, deceleration and ultrasound score. The sensitivity, specificity, percentage of support vectors and total classification error of this set were respectively 100%, 100%, 35% and 0%. This indicated our ability to separate the data into two sets (normal fetuses and pathologic fetuses), and the results highlight the excellent match with the clinical classification performed by the physicians. This work indicates the feasibility of detecting compromised fetuses and also represents an interesting method of close fetal monitoring during the entire pregnancy.

n-Order and maximum fuzzy similarity entropy for discrimination of signals of different complexity: Application to fetal heart rate signals.

This paper presents two new concepts for discrimination of signals of different complexity. The first focused initially on solving the problem of setting entropy descriptors by varying the pattern size instead of the tolerance. This led to the search for the optimal pattern size that maximized the similarity entropy. The second paradigm was based on the n-order similarity entropy that encompasses the 1-order similarity entropy. To improve the statistical stability, n-order fuzzy similarity entropy was proposed. Fractional Brownian motion was simulated to validate the different methods proposed, and fetal heart rate signals were used to discriminate normal from abnormal fetuses. In all cases, it was found that it was possible to discriminate time series of different complexity such as fractional Brownian motion and fetal heart rate signals. The best levels of performance in terms of sensitivity (90%) and specificity (90%) were obtained with the n-order fuzzy similarity entropy. However, it was shown that the optimal pattern size and the maximum similarity measurement were related to intrinsic features of the time series.

High resolution processing techniques for ultrasound doppler velocimetry in the presence of colored noise. Part I: Nonstationary methods.

Real-time flow velocity measurement is a practical issue in industrial and biomedical applications. Because their good frequency resolution, parametric methods such as autoregressive (AR) modeling and time-frequency distributions (TFD) are generally preferred to Fourier analysis. However, these methods become highly inaccurate in the presence of colored noise. We review here the principal parametric and nonparametric techniques and show their limitations in the estimation of Doppler frequency in the presence of strong colored noise. Different solutions to overcome these limitations are then proposed and compared using synthetic Doppler signals with colored noise.

High resolution processing techniques for ultrasound doppler velocimetry in the presence of colored noise. Part II: Multiplephase pipe-flow velocity measurement.

This paper presents an application of continuous wave ultrasound Doppler velocity measurements to two-phase flow in pipes. In many petroleum wells, the multiphase flow is separated into two phases: the first is a liquid phase and the second is a gas phase with small scatterers. The problem of multiphase velocity profile measurements has not been satisfactorily solved by classical approaches due to the multiphase nature of the fluid and the presence of colored noise, which introduces a significant bias in classical frequency estimators. We propose the use of high resolution frequency techniques to overcome the classical limitations. Direct estimation of Doppler frequency then obtained using either time frequency maximum frequency or arguments of poles of the parametric model that identifies the Doppler part of the signal is discussed. The tests made with synthetic Doppler signals and two-phase flow have demonstrated the excellent performance of the high resolution techniques based on reassignment and parametric techniques.

New estimators and guidelines for better use of fetal heart rate estimators with Doppler ultrasound devices.

Characterizing fetal wellbeing with a Doppler ultrasound device requires computation of a score based on fetal parameters. In order to analyze the parameters derived from the fetal heart rate correctly, an accuracy of 0.25 beats per minute is needed. Simultaneously with the lowest false negative rate and the highest sensitivity, we investigated whether various Doppler techniques ensure this accuracy. We found that the accuracy was ensured if directional Doppler signals and autocorrelation estimation were used. Our best estimator provided sensitivity of 95.5%, corresponding to an improvement of 14% compared to the standard estimator.

Automatic optimal input command for linearization of cMUT output by a temporal target.

Capacitive micromachined ultrasonic transducers (cMUTs) are a promising alternative to the piezoelectric transducer. However, their native nonlinear behavior is a limitation for their use in medical ultrasound applications. Several methods based on the pre-compensation of a preselected input voltage have been proposed to cancel out the harmonic components generated. Unfortunately, these existing pre-compensation methods have two major flaws. The first is that the pre-compensation procedure is not generally automatic, and the second is that they can only reduce the second harmonic component. This can, therefore, limit their use for some imaging methods, which require a broader bandwidth, e.g., to receive the third harmonic component. In this study, we generalized the presetting methods to reduce all nonlinearities in the cMUT output. Our automatic pre-compensation method can work whatever the excitation waveform. The precompensation method is based on the nonlinear modeling of harmonic components from a Volterra decomposition in which the parameters are evaluated by using a Nelder-Mead algorithm. To validate the feasibility of this approach, the method was applied to an element of a linear array with several types of excitation often encountered in encoded ultrasound imaging. The results showed that the nonlinear components were reduced by up to 21.2 dB.

Optimization of contrast-to-tissue ratio by adaptation of transmitted ternary signal in ultrasound pulse inversion imaging.

Ultrasound contrast imaging has provided more accurate medical diagnoses thanks to the development of innovating modalities like the pulse inversion imaging. However, this latter modality that improves the contrast-to-tissue ratio (CTR) is not optimal, since the frequency is manually chosen jointly with the probe. However, an optimal choice of this command is possible, but it requires precise information about the transducer and the medium which can be experimentally difficult to obtain, even inaccessible. It turns out that the optimization can become more complex by taking into account the kind of generators, since the generators of electrical signals in a conventional ultrasound scanner can be unipolar, bipolar, or tripolar. Our aim was to seek the ternary command which maximized the CTR. By combining a genetic algorithm and a closed loop, the system automatically proposed the optimal ternary command. In simulation, the gain compared with the usual ternary signal could reach about 3.9 dB. Another interesting finding was that, in contrast to what is generally accepted, the optimal command was not a fixed-frequency signal but had harmonic components.