Hidden Markov model in nonnegative matrix factorization for fetal heart rate estimation using physiological priors.

Objective.Fetal heart rate (fHR) analysis remains the most common technique for detecting fetal distress when monitoring the fetal well-being during labor. If cardiotocography (CTG) is nowadays the non-invasive clinical reference technique for fHR measurement, it suffers from several drawbacks, hence an increasing interest towards alternative technologies, especially around abdominal ECG (aECG).Approach.An original solution, using a single abdominal lead, was recently proposed to address both the feasibility in clinical routine and the challenging detection of temporal events when facing interfered signals from real life conditions. Based on a specification of the non-negative matrix factorization (NMF) algorithm, it exploits the semi-periodicity of fetal electrocardiogram (fECG) for fHR estimation. However, this method assumes temporal independence and therefore does not consider the continuity property of fHR values. It is thus proposed to add to the NMF framework a hidden Markov model (HMM) to include physiological information about fHR temporal evolution. Under a statistical setting, constraints have been added by accommodating regularization terms through Bayesian priors.Main results.The proposed method is evaluated on 23 real aECG signals from a new clinical database, according to CTG reference, and compared with the original NMF-only algorithm. The new proposed method improves performance, with an agreement with CTG increasing from 71% to 80%.Significance.This highlights the interest of a better modelization of the fHR characteristics for a more robust estimation.

Residual convolutional autoencoder combined with a non-negative matrix factorization to estimate fetal heart rate.

The fetal heart rate (fHR) plays an important role in the determination of the good health of the fetus. Beside the traditional Doppler ultrasound technique, non-invasive fetal electrocardiography (fECG) has become an interesting alternative. However, extracting clean fECG from abdominal ECG (aECG) recordings is a challenging task due to the presence of the maternal ECG component and various noise sources. In this context, we propose a deep residual convolutional autoencoder network trained on synthetic aECG simulations followed by a transfer learning phase on real aECG recordings to extract the cleanest fECG. Afterwards, we propose to use a non-negative matrix factorization based approach on the obtained fECG to estimate the fHR. Our method is evaluated on three publicly available databases demonstrating that it can provide significant performance improvement against comparative methodologies. Clinical relevance- The presented method has the advantage of estimating the fetal heart rate from a single-channel abdominal electrocardiogram without prior knowledge on the noise sources nor the maternal R-peak locations.

Fetal ECG denoising using dynamic time warping template subtraction.

The analysis of the fetal electrocardiogram (ECG) requires to remove the mother ECG (mECG) from the abdominal ECG signals. Template subtraction is a method that consists in modeling and removing the mECG's mean period i.e. the signal waveform defined as the Euclidean mean of all periods. This mean period is then subtracted to all periods to extract the fetal ECG (fECG). Such a method is not accurate because each mECG's period is not correctly aligned with the mean period. We propose to take account of the diffeomorphism of each period to improve the precision of the model and remove the mECG more efficiently. The soft-dynamic time warping (DTW) algorithm is used to compute the mean mECG period and the alignment between the mean period and all periods. Our approach is compared to a classic template subtraction on synthetic and real databases. Results show that considering the dynamic time warping allows a better removal of the mECG. Clinical relevance - The template subtraction is modified in this paper to consider the time warping for each mother ECG's period in order to improve the fetal ECG extraction from the abdominal ECG.

Comparison of Blind Source Separation Methods to Surface Electromyogram for Extensor Muscles of the Index and Little Fingers.

Crosstalk is the result of the propagation of muscle electrical signals on surface electromyogram channels simultaneously. The objective of this paper is to study the behavior of three blind source separation (BSS) methods for crosstalk reduction during finger extensor muscle contractions: FastICA, joint diagonalization of covariance matrices and optimal filtering. These methods have been tested on artificial mixtures defined by a temporal sum of the real signals from isolated contraction of two independent biomechanical muscles for the extension of the index and little finger. Artificial mixtures display a ground truth for comparison between the methods. The separation was better using the optimal filtering compared to the other two methods. The optimal filtering have then be tested on real mixtures recorded during a simultaneous contraction of the two muscles. The results are less satisfactory but open doors to new perspectives.

U-Net Neural Network for Heartbeat Detection in Ballistocardiography.

Monitoring vital signs of neonates can be harmful and lead to developmental troubles. Ballistocardiography, a contactless heart rate monitoring method, has the potential to reduce this monitoring pain. However, signal processing is uneasy due to noise, inherent physiological variability and artifacts (e.g. respiratory amplitude modulation and body position shifts). We propose a new heartbeat detection method using neural networks to learn this variability. A U-Net model takes thirty-second-long records as inputs and acts like a nonlinear filter. For each record, it outputs the samples probabilities of belonging to IJK segments. A heartbeat detection algorithm finally detects heartbeats from those segments, based on a distance criterion. The U-Net has been trained on 30 healthy subjects and tested on 10 healthy subjects, from 8 to 74 years old. Heartbeats have been detected with 92% precision and 80% recall, with possible optimization in the future to achieve better performance.

Is there an Optimal Localization of Cardio-microphone Sensors for Phonocardiogram Analysis?

Heart auscultation is one of the most useful medical diagnostic tools for getting valuable information of heart valves and heart hemodynamics functions. However, the information acquired by a traditional stethoscope can be inaccurate and insufficient. Phonocardiogram (PCG) was developed to improve accuracy through visual inspection and analysis. Digitally processed, PCG can then be analyzed by automated heart sound analysis systems. But there is no standardization for PCG data acquisition unlike electrocardiogram (ECG). This study aims at analyzing the influence of cardiomicrophone localization on the chest for the study of cardiac sounds S1 and S2. For that purpose, simultaneous acquisitions of 12 PCG signals with one ECG signal were realized and a comparative analysis of delays between R waves of ECG and detected S1 and S2 sounds was conducted. Results show that there are significant differences between R-S1 (or R-S2) intervals obtained from different areas of sensor placement on the chest. For future works on PCG, studies dealing with the analysis of heart sounds or proposing new heart sounds detection algorithms may pay attention to the location and attachment of PCG sensors.

Fetal heart rate estimation from a single phonocardiogram signal using non-negative matrix factorization.

Fetal heart rate (FHR) is an important characteristic in fetal well-being follow-up. It is classically estimated using the cardiotocogram (CTG) non-invasive reference technique. However, this latter presents some significant drawbacks. An alternative non-invasive solution based on the fetal phonocardiogram (fetal PCG) can be used. But most of proposed methods based on the PCG signal need to detect and to label the fetal cardiac S1 and S2 sounds, which may be a difficult task in certain conditions. Therefore, in this paper, we propose a new methodology for FHR estimation from fetal PCG with one single cardio-microphone and without the distinction constraint of heart sounds. The method is based on the non-negative matrix factorization (NMF) applied on the spectrogram of fetal PCG considered as a source-filter model. The proposed method provides satisfactory results on a preliminary dataset of abdominal PCG signals. When compared to the reference CTG, correlation on FHR estimations between PCG and CTG is around 90%.

Temporal Dynamics of Natural Static Emotional Facial Expressions Decoding: A Study Using Event- and Eye Fixation-Related Potentials.

This study aims at examining the precise temporal dynamics of the emotional facial decoding as it unfolds in the brain, according to the emotions displayed. To characterize this processing as it occurs in ecological settings, we focused on unconstrained visual explorations of natural emotional faces (i.e., free eye movements). The General Linear Model (GLM; Smith and Kutas, 2015a,b; Kristensen et al., 2017a) enables such a depiction. It allows deconvolving adjacent overlapping responses of the eye fixation-related potentials (EFRPs) elicited by the subsequent fixations and the event-related potentials (ERPs) elicited at the stimuli onset. Nineteen participants were displayed with spontaneous static facial expressions of emotions (Neutral, Disgust, Surprise, and Happiness) from the DynEmo database (Tcherkassof et al., 2013). Behavioral results on participants' eye movements show that the usual diagnostic features in emotional decoding (eyes for negative facial displays and mouth for positive ones) are consistent with the literature. The impact of emotional category on both the ERPs and the EFRPs elicited by the free exploration of the emotional faces is observed upon the temporal dynamics of the emotional facial expression processing. Regarding the ERP at stimulus onset, there is a significant emotion-dependent modulation of the P2-P3 complex and LPP components' amplitude at the left frontal site for the ERPs computed by averaging. Yet, the GLM reveals the impact of subsequent fixations on the ERPs time-locked on stimulus onset. Results are also in line with the valence hypothesis. The observed differences between the two estimation methods (Average vs. GLM) suggest the predominance of the right hemisphere at the stimulus onset and the implication of the left hemisphere in the processing of the information encoded by subsequent fixations. Concerning the first EFRP, the Lambda response and the P2 component are modulated by the emotion of surprise compared to the neutral emotion, suggesting an impact of high-level factors, in parieto-occipital sites. Moreover, no difference is observed on the second and subsequent EFRP. Taken together, the results stress the significant gain obtained in analyzing the EFRPs using the GLM method and pave the way toward efficient ecological emotional dynamic stimuli analyses.

Regularization and a general linear model for event-related potential estimation.

The usual event-related potential (ERP) estimation is the average across epochs time-locked on stimuli of interest. These stimuli are repeated several times to improve the signal-to-noise ratio (SNR) and only one evoked potential is estimated inside the temporal window of interest. Consequently, the average estimation does not take into account other neural responses within the same epoch that are due to short inter stimuli intervals. These adjacent neural responses may overlap and distort the evoked potential of interest. This overlapping process is a significant issue for the eye fixation-related potential (EFRP) technique in which the epochs are time-locked on the ocular fixations. The inter fixation intervals are not experimentally controlled and can be shorter than the neural response's latency. To begin, the Tikhonov regularization, applied to the classical average estimation, was introduced to improve the SNR for a given number of trials. The generalized cross validation was chosen to obtain the optimal value of the ridge parameter. Then, to deal with the issue of overlapping, the general linear model (GLM), was used to extract all neural responses inside an epoch. Finally, the regularization was also applied to it. The models (the classical average and the GLM with and without regularization) were compared on both simulated data and real datasets from a visual scene exploration in co-registration with an eye-tracker, and from a P300 Speller experiment. The regularization was found to improve the estimation by average for a given number of trials. The GLM was more robust and efficient, its efficiency actually reinforced by the regularization.

Estimation of overlapped Eye Fixation Related Potentials: The General Linear Model, a more flexible framework than the ADJAR algorithm.

The Eye Fixation Related Potential (EFRP) estimation is the average of EEG signals across epochs at ocular fixation onset. Its main limitation is the overlapping issue. Inter Fixation Intervals (IFI) - typically around 300 ms in the case of unrestricted eye movement- depend on participants' oculomotor patterns, and can be shorter than the latency of the components of the evoked potential. If the duration of an epoch is longer than the IFI value, more than one fixation can occur, and some overlapping between adjacent neural responses ensues. The classical average does not take into account either the presence of several fixations during an epoch or overlapping. The Adjacent Response algorithm (ADJAR), which is popular for event-related potential estimation, was compared to the General Linear Model (GLM) on a real dataset from a conjoint EEG and eye-tracking experiment to address the overlapping issue. The results showed that the ADJAR algorithm was based on assumptions that were too restrictive for EFRP estimation. The General Linear Model appeared to be more robust and efficient. Different configurations of this model were compared to estimate the potential elicited at image onset, as well as EFRP at the beginning of exploration. These configurations took into account the overlap between the event-related potential at stimulus presentation and the following EFRP, and the distinction between the potential elicited by the first fixation onset and subsequent ones. The choice of the General Linear Model configuration was a tradeoff between assumptions about expected behavior and the quality of the EFRP estimation: the number of different potentials estimated by a given model must be controlled to avoid erroneous estimations with large variances.

Feasibility of BCI Control in a Realistic Smart Home Environment.

Smart homes have been an active area of research, however despite considerable investment, they are not yet a reality for end-users. Moreover, there are still accessibility challenges for the elderly or the disabled, two of the main potential targets for home automation. In this exploratory study we design a control mechanism for smart homes based on Brain Computer Interfaces (BCI) and apply it in the "Domus" smart home platform in order to evaluate the potential interest of users about BCIs at home. We enable users to control lighting, a TV set, a coffee machine and the shutters of the smart home. We evaluate the performance (accuracy, interaction time), usability and feasibility (USE questionnaire) on 12 healthy subjects and 2 disabled subjects. We find that healthy subjects achieve 77% task accuracy. However, disabled subjects achieved a better accuracy (81% compared to 77%).

Multimodal approach to estimate the ocular movements during EEG recordings: A coupled tensor factorization method.

This paper deals with coupled tensor factorization. A relaxed criterion derived from the advanced coupled matrix-tensor factorization (ACMTF) proposed by Acar et al. is described. The proposed relaxed ACMTF (RACMTF) criterion is based on weaker assumptions that are thus more often satisfied when dealing with actual data. Numerical simulations show the benefit of using jointly two data sets when the underlying factors are highly correlated, especially if one of the modality is less noisy than the other one. The proposed method is finally applied on actual Gaze&EEG data to estimate the ocular artifacts into the EEG recordings.

An application of Gaussian processes on ocular artifact removal from EEG.

Consequences of eye movements are one of the main interferences that distort the brain EEG recordings. In this paper, a multi-modal approach is used to estimate the ocular artifacts in the EEG: both vertical and horizontal eye movement signals recorded by an eye tracker are used as a reference to denoise the EEG. A Gaussian process, i.e. a second order statistics method, is assumed to model the link between the eye tracker signals and the EEG signals. The proposed method is thus a non-linear extension of the well-known adaptive filtering and can be applied with a single EEG signal contrary to independent component analysis (ICA) which is extensively used. The results show the applicability and the efficiency of this model on the ocular artifact removal.

Correction: cecotti, h. And rivet, B. Subject combination and electrode selection in cooperative brain-computer interface based on event related potentials. Brain sci. 2014, 4, 335-355.

The authors wish to make the following correction to this paper (Cecotti, H.; Rivet, B. Subject Combination and Electrode Selection in Cooperative Brain-Computer Interface Based on Event Related Potentials. Brain Sci. 2014, 4, 335-355). Dut to an error the reference number in the original published paper were not shown. The former main text should be replaced as below.

Subject combination and electrode selection in cooperative brain-computer interface based on event related potentials.

New paradigms are required in Brain-Computer Interface (BCI) systems for the needs and expectations of healthy people. To solve this issue, we explore the emerging field of cooperative BCIs, which involves several users in a single BCI system. Contrary to classical BCIs that are dependent on the unique subject's will, cooperative BCIs are used for problem solving tasks where several people shall be engaged by sharing a common goal. Similarly as combining trials over time improves performance, combining trials across subjects can significantly improve performance compared with when only a single user is involved. Yet, cooperative BCIs may only be used in particular settings, and new paradigms must be proposed to efficiently use this approach. The possible benefits of using several subjects are addressed, and compared with current single-subject BCI paradigms. To show the advantages of a cooperative BCI, we evaluate the performance of combining decisions across subjects with data from an event-related potentials (ERP) based experiment where each subject observed the same sequence of visual stimuli. Furthermore, we show that it is possible to achieve a mean AUC superior to 0.95 with 10 subjects and 3 electrodes on each subject, or with 4 subjects and 6 electrodes on each subject. Several emerging challenges and possible applications are proposed to highlight how cooperative BCIs could be efficiently used with current technologies and leverage BCI applications.

Modeling quasi-periodic signals by a non-parametric model: application on fetal ECG extraction.

Quasi-periodic signals can be modeled by their second order statistics as Gaussian process. This work presents a non-parametric method to model such signals. ECG, as a quasi-periodic signal, can also be modeled by such method which can help to extract the fetal ECG from the maternal ECG signal, using a single source abdominal channel. The prior information on the signal shape, and on the maternal and fetal RR interval, helps to better estimate the parameters while applying the Bayesian principles. The values of the parameters of the method, among which the R-peak instants, are accurately estimated using the Metropolis-Hastings algorithm. This estimation provides very precise values for the R-peaks, so that they can be located even between the existing time samples.

Fetal ECG extraction by extended state Kalman filtering based on single-channel recordings.

In this paper, we present an extended nonlinear Bayesian filtering framework for extracting electrocardiograms (ECGs) from a single channel as encountered in the fetal ECG extraction from abdominal sensor. The recorded signals are modeled as the summation of several ECGs. Each of them is described by a nonlinear dynamic model, previously presented for the generation of a highly realistic synthetic ECG. Consequently, each ECG has a corresponding term in this model and can thus be efficiently discriminated even if the waves overlap in time. The parameter sensitivity analysis for different values of noise level, amplitude, and heart rate ratios between fetal and maternal ECGs shows its effectiveness for a large set of values of these parameters. This framework is also validated on the extractions of fetal ECG from actual abdominal recordings, as well as of actual twin magnetocardiograms.

Adaptive training session for a P300 speller brain-computer interface.

With a brain-computer interface (BCI), it is nowadays possible to achieve a direct pathway between the brain and computers thanks to the analysis of some particular brain activities. The detection of even-related potentials, like the P300 in the oddball paradigm exploited in P300-speller, provides a way to create BCIs by assigning several detected ERP to a command. Due to the noise present in the electroencephalographic signal, the detection of an ERP and its different components requires efficient signal processing and machine learning techniques. As a consequence, a calibration session is needed for training the models, which can be a drawback if its duration is too long. Although the model depends on the subject, the goal is to provide a reliable model for the P300 detection over time. In this study, we propose a new method to evaluate the optimal number of symbols (i.e. the number of ERP that shall be detected given a determined target probability) that should be spelt during the calibration process. The goal is to provide a usable system with a minimum calibration duration and such that it can automatically switch between the training and online sessions. The method allows to adaptively adjust the number of training symbols to each subject. The evaluation has been tested on data recorded on 20 healthy subjects. This procedure lets drastically reduced the calibration session: height symbols during the training session reach an initialized system with an average accuracy of 80% after five epochs.

EEG sensor selection by sparse spatial filtering in P300 speller brain-computer interface.

A Brain-Computer Interface (BCI) is a specific type of human-machine interface that enables communication between a subject/patient and a computer by direct control from decoding of brain activity. This paper deals with the P300-speller application that enables to write a text based on the oddball paradigm. To improve the ergonomics and minimize the cost of such a BCI, reducing the number of electrodes is mandatory. We propose a new algorithm to select a relevant subset of electrodes by estimating sparse spatial filters. A l(1)-norm penalization term, as an approximation of the l(0)-norm, is introduced in the xDAWN algorithm, which maximizes the signal to signal-plus-noise ratio. Experimental results on 20 subjects show that the proposed method is efficient to select the most relevant sensors: from 32 down to 10 sensors, the loss in classification accuracy is less than 5%.

A study of lip movements during spontaneous dialog and its application to voice activity detection.

This paper presents a quantitative and comprehensive study of the lip movements of a given speaker in different speech/nonspeech contexts, with a particular focus on silences (i.e., when no sound is produced by the speaker). The aim is to characterize the relationship between "lip activity" and "speech activity" and then to use visual speech information as a voice activity detector (VAD). To this aim, an original audiovisual corpus was recorded with two speakers involved in a face-to-face spontaneous dialog, although being in separate rooms. Each speaker communicated with the other using a microphone, a camera, a screen, and headphones. This system was used to capture separate audio stimuli for each speaker and to synchronously monitor the speaker's lip movements. A comprehensive analysis was carried out on the lip shapes and lip movements in either silence or nonsilence (i.e., speech+nonspeech audible events). A single visual parameter, defined to characterize the lip movements, was shown to be efficient for the detection of silence sections. This results in a visual VAD that can be used in any kind of environment noise, including intricate and highly nonstationary noises, e.g., multiple and/or moving noise sources or competing speech signals.