Notes

Likelihood of emergency solutions associated with surrounding heat, great air particle and its elements.
The classifier with best performance (accuracy = 0.88, precision = 0.91, recall = 0.87, specificity = 0.91, F1-score = 0.81) was trained with features extracted from the combination of early inspiratory phase and entire inspiratory phase. To our best knowledge, we are the first to address the challenging multi-classification problem.Tracheal sounds represent information about the upper airway and respiratory airflow, however, they can be contaminated by the snoring sounds. The sound of snoring has spectral content in a wide range that overlaps with that of breathing sounds during sleep. For assessing respiratory airflow using tracheal breathing sound, it is essential to remove the effect of snoring. In this paper, an automatic and unsupervised wavelet-based snoring removal algorithm is presented. Simultaneously with full-night polysomnography, the tracheal sound signals of 9 subjects with different levels of airway obstruction were recorded by a microphone placed over the trachea during sleep. The segments of tracheal sounds that were contaminated by snoring were manually identified through listening to the recordings. The selected segments were automatically categorized based on including discrete or continuous snoring pattern. Segments with discrete snoring were analyzed by an iterative wave-based filtering optimized to separate large spectral components related to snoring from smaller ones corresponded to breathing. Those with continuous snoring were first segmented into shorter segments. Then, each short segments were similarly analyzed along with a segment of normal breathing extracted from the recordings during wakefulness. The algorithm was evaluated by visual inspection of the denoised sound energy and comparison of the spectral densities before and after removing snores, where the overall rate of detectability of snoring was less than 2%.Clinical Relevance- The algorithm provides a way of separating snoring pattern from the tracheal breathing sounds. Therefore, each of them can be analyzed separately to assess respiratory airflow and the pathophysiology of the upper airway during sleep.We propose a robust and efficient lung sound classification system using a snapshot ensemble of convolutional neural networks (CNNs). A robust CNN architecture is used to extract high-level features from log mel spectrograms. The CNN architecture is trained on a cosine cycle learning rate schedule. Capturing the best model of each training cycle allows to obtain multiple models settled on various local optima from cycle to cycle at the cost of training a single mode. Therefore, the snapshot ensemble boosts performance of the proposed system while keeping the drawback of expensive training of ensembles moderate. To deal with the class-imbalance of the dataset, temporal stretching and vocal tract length perturbation (VTLP) for data augmentation and the focal loss objective are used. Empirically, our system outperforms state-of-the-art systems for the prediction task of four classes (normal, crackles, wheezes, and both crackles and wheezes) and two classes (normal and abnormal (i.e. crackles, wheezes, and both crackles and wheezes)) and achieves 78.4% and 83.7% ICBHI specific micro-averaged accuracy, respectively. The average accuracy is repeated on ten random splittings of 80% training and 20% testing data using the ICBHI 2017 dataset of respiratory cycles.This paper focuses on the use of an attention-based encoder-decoder model for the task of breathing sound segmentation and detection. This study aims to accurately segment the inspiration and expiration of patients with pulmonary diseases using the proposed model. Spectrograms of the lung sound signals and labels for every time segment were used to train the model. The model would first encode the spectrogram and then detect inspiratory or expiratory sounds using the encoded image on an attention-based decoder. Physicians would be able to make a more precise diagnosis based on the more interpretable outputs with the assistance of the attention mechanism.The respiratory sounds used for training and testing were recorded from 22 participants using digital stethoscopes or anti-noising microphone sets. Experimental results showed a high 92.006% accuracy when applied 0.5 second time segments and ResNet101 as encoder. Consistent performance of the proposed method can be observed from ten-fold cross-validation experiments.In addition to the global parameter- and time-series-based approaches, physiological analyses should constitute a local temporal one, particularly when analyzing data within protocol segments. Hence, we introduce the R package implementing the estimation of temporal orders with a causal vector (CV). It may use linear modeling or time series distance. The algorithm was tested on cardiorespiratory data comprising tidal volume and tachogram curves, obtained from elite athletes (supine and standing, in static conditions) and a control group (different rates and depths of breathing, while supine). We checked the relation between CV and body position or breathing style. The rate of breathing had a greater impact on the CV than does the depth. The tachogram curve preceded the tidal volume relatively more when breathing was slower.The recent progress in recognizing low-resolution instantaneous high-density surface electromyography (HD-sEMG) images opens up new avenues for the development of more fluid and natural muscle-computer interfaces. However, the existing approaches employed a very large deep convolutional neural network (ConvNet) architecture and complex training schemes for HD-sEMG image recognition, which requires learning of >5.63 million(M) training parameters only during fine-tuning and pre-trained on a very large-scale labeled HD-sEMG training dataset, as a result, it makes high-end resource-bounded and computationally expensive. To overcome this problem, we propose S-ConvNet models, a simple yet efficient framework for learning instantaneous HD-sEMG images from scratch using random-initialization. Without using any pre-trained models, our proposed S-ConvNet demonstrate very competitive recognition accuracy to the more complex state of the art, while reducing learning parameters to only ≈ 2M and using ≈ 12 × smaller dataset. The experimental results proved that the proposed S-ConvNet is highly effective for learning discriminative features for instantaneous HD-sEMG image recognition, especially in the data and high-end resource-constrained scenarios.Modeling of surface electromyographic (EMG) signal has been proven valuable for signal interpretation and algorithm validation. However, most EMG models are currently limited to single muscle, either with numerical or analytical approaches. Here, we present a preliminary study of a subject-specific EMG model with multiple muscles. Magnetic resonance (MR) technique is used to acquire accurate cross section of the upper limb and contours of five muscle heads (biceps brachii, brachialis, lateral head, medial head, and long head of triceps brachii). The MR image is adjusted to an idealized cylindrical volume conductor model by image registration. High-density surface EMG signals are generated for two movements - elbow flexion and elbow extension. The simulated and experimental potentials were compared using activation maps. Similar activation zones were observed for each movement. These preliminary results indicate the feasibility of the multi-muscle model to generate EMG signals for complex movements, thus providing reliable data for algorithm validation.In the last decade, accurate identification of motor unit (MU) firings received a lot of research interest. Different decomposition methods have been developed, each with its advantages and disadvantages. In this study, we evaluated the capability of three different types of neural networks (NNs), namely dense NN, long short-term memory (LSTM) NN and convolutional NN, to identify MU firings from high-density surface electromyograms (HDsEMG). Each type of NN was evaluated on simulated HDsEMG signals with a known MU firing pattern and high variety of MU characteristics. Compared to dense NN, LSTM and convolutional NN yielded significantly higher precision and significantly lower miss rate of MU identification. LSTM NN demonstrated higher sensitivity to noise than convolutional NN.Clinical Relevance-MU identification from HDsEMG signals offers valuable insight into neurophysiology of motor system but requires relatively high level of expert knowledge. This study assesses the capability of self-learning artificial neural networks to cope with this problem.In this study, an attempt has been made to distinguish between nonfatigue and fatigue conditions in surface Electromyography (sEMG) signal using the time frequency distribution obtained from analytic Bump Continuous Wavelet Transform. For the analysis, sEMG signals from biceps brachii muscle of 22 healthy subjects are acquired during isometric contraction protocol. The signals acquired is preprocessed and partitioned into ten equal segments followed by the decomposition of selected segments using analytic Bump wavelets. Further, Singular Value Decomposition is applied to the time frequency distribution matrix and the maximum singular value and entropy feature for each segment are obtained. The usefulness of both the features is estimated using the Wilcoxon sign rank test that gives higher significance with a p less then .00001. It is observed that the proposed method is capable of analyzing the fatigue regions in sEMG signals.Surface electromyogram (sEMG) has been widely applied in neurorehabilitation techniques such as human-machine interface (HMI). The individual difference of sEMG characteristics has long been a challenge for multi-user HMI. However, the individually unique sEMG property indicates its high potential as a biometrics modality. In this work, we propose a novel application of high-density sEMG (HD-sEMG) for personal identification. HD-sEMG can decode the high-resolution spatial patterns of muscle activations, besides the widely studied temporal features, thus providing more sufficient information. We acquired 64-channel HD-sEMG signals on the dorsum of the right hand from 22 subjects during finger muscle isometric contractions. We achieved an accuracy of 99.5% to recognize the identity of each subject, demonstrating the excellent performance of HD-sEMG for personal identification. To the best of our knowledge, this is the first study to employ HD-sEMG for personal identification.Clinical relevance-Our work has proved the huge individual difference of HD-sEMG, which may result from the individually unique bioelectrophysiological activity of human body, deriving from both neural and biomechanical factors. The investigation of subject-specific HD-sEMG pattern may contribute to a better design of subject-specific clinical rehabilitation robots and a deeper understanding of human movement mechanism.Electromyography offers a way to interface an amputee's resilient muscles to control a bionic prosthesis. While myoelectric prostheses are promising, user acceptance of these devices remain low due to a lack of intuitiveness and ease-of-use. Using a low-cost wearable flexible electrodes array, the proposed system leverages high-density surface electromyography (HD-EMG) and deep learning techniques to classify forearm muscle contractions. These techniques allow for increased intuitiveness and ease-of-use of a myoelectric control scheme with a single easy-to-install electrodes apparatus. This paper proposes a flexible electrodes array construction using standard printed circuit board manufacturing processes for low-cost and quick design-to-production cycles. HD-EMG dataset visualization with t-distributed Stochastic Neighbor Embedding (t-SNE) is introduced, and offline classification results of the wearable gesture recognition system for hand prosthesis control are validated on a group of 8 able-bodied subjects. Using a majority vote on 5 successive inferences, a median recognition accuracy of 98.61 % was obtained across the group for an 8 gestures set. For a 6 gestures set containing commonly used prosthesis positions, the median accuracy reached 99.57 % with the majority vote.In this study, the feasibility of conducting a concurrent estimation of drowsiness, stress, and tiredness by heart rate variability (HRV) in a driving simulator environment was examined. Subjects were required to attend a 120-min driving session four times two morning and two afternoon sessions. Blood pressure and salivary amylase were also recorded to assess acute stress. A set of estimators was prepared, and stepwise regression was conducted on two different models at p = 0.05. In this work, it was shown that the use of a stepwise method and additional estimator capable of extracting significant and relevant information for multiple emotions with average performance in the form of the correlation coefficient(root mean square error) can increase up to 0.68 ± 0.12 (0.66 ± 0.28), 0.72 ± 0.13 (0.43 ± 0.21), and 0.71 ± 0.13 (0.48 ± 0.21), corresponding to drowsiness, stress, and tiredness, respectively. The results suggest that a single time series of HRV can extract more than one emotion, thus enabling a concurrent model to be developed. It was also observed that physiological behavior while driving works in a more complex way. The current evidence indicates the feasibility of conducting concurrent emotion assessment during driving.Early and noninvasive identification of heart failure progression is an important adjunct to successful and timely intervention. Severity of heart failure (HF) was assessed by Left Ventricular Ejection Fraction (LVEF). In this paper, we explore the circadian (24-hour) heart rate variability (HRV) features from ''normal" (EF >50%), "at-risk" (EF less then 40%), and "border-line" (40% ≤ EF ≤ 50%) patient data to determine whether HRV features can predict the stage of heart failure. All coronary artery disease (CAD) 24-hour circadian heart rate data were fitted by a cosinor analysis algorithm. Hourly HRV features from time- and frequency-domains were then extracted from all 24-hour patient data. A one-way ANOVA test was performed followed by a Tukey post-hoc multiple comparison test to investigate the differences among the three groups. The results showed a statistically significant difference between the three groups when using the normalized high frequency (HF Norm), low frequency peak (LF Peak), and the normalized very-low frequency (VLF Norm) for the 0500-0600 and 1800-1900 time periods. These results highlight a possible link between the circadian variation of sympathetic and parasympathetic nervous system activity and LVEF for CAD patients. The results could be useful in differentiating the various degrees of LVEF by using only noninvasive HRV features derived over a 24-hour period.Clinical relevance- The proposed method could be clinically useful to estimate the extent of LVEF associated with the severity of heart failure by recording the circadian variation of the heart rate in CAD patients. However, further clinical trials on a larger cohort of patients and controls are required.Recent developments of detrended fluctuation analysis (DFA) provide multifractal/multiscale (MFMS) descriptions of the heart rate self-similarity, a promising approach to cardiovascular complexity. However, it is unclear whether the MFMS DFA may also describe the nonlinear components of heart rate variability. Our aim is to define MFMS DFA indices for quantifying the short-term and long-term degree of the heart-rate nonlinearity and to apply these indices to detect possible sex-related differences.We recorded the inter-beat-interval (IBI) series in 42 male and in 42 female healthy participants sitting at rest for about 2 hours. For each series j, we generated 100 phase-randomized surrogate series. We applied the MFMS DFA to estimate the self-similarity coefficients α over scales τ between 8 and 512 s and moment orders q between -5 and +5, obtaining coefficients for the original series, αO,j (q, τ), and for each surrogate, αi,j (q, τ) with 1≤i≤100. We first evaluated πj(q, τ), percentile of αi,j (q, τ) distribution in which was αO,j (q, τ). Then we calculated the percentages of scales where πj(q, τ) was less then 5% for 8≤τ≤16 s (short-term nonlinearity index NL1(q)) and for 16≤τ≤512 s (long-term nonlinearity index NL2(q)). We found that NL1(q) was generally greater than 50% at all q≥0 but q=2 (i.e., moment order of the monofractal DFA), while at q less then 0 it was high in males only, with significant sex differences at q=-1 and q=-2. Results indicate that the multifractal DFA may highlight nonlinear heart-rate components at the short scales that are not revealed by the traditional monofractal DFA and that appear related to gender differences.Clinical Relevance- This supports the use of MFMS DFA to integrate the linear information from traditional spectral methods of heart rate variability in clinical studies aimed at improving the stratification of the cardiovascular risk.Heart rate variability (HRV) measures the regularity between consecutive heartbeats driven by the balance between the sympathetic and parasympathetic branches of the autonomous nervous system. Wearable devices embedding photoplethysmogram (PPG) technology can be used to derive HRV, creating many opportunities for remote monitoring of this physiological parameter. However, uncontrolled conditions met in daily life pose several challenges related to disturbances that can deteriorate the PPG signal, making the calculation of HRV metrics untrustworthy and not reliable. In this work, we propose a HRV quality metric that is directly related to the HRV accuracy and can be used to distinguish between accurate and inaccurate HRV values. A parametric supervised approach estimates HRV accuracy using a model whose inputs are features extracted from the PPG signal and the output is the HRV error between HRV metrics obtained from the PPG and the ECG collected during an experimental protocol involving several activities. The estimated HRV accuracy of the model is used as an indication of the HRV quality.Diverse analysis techniques have been used to comprehend the regulation by the autonomic nervous system (ANS) of the cardiovascular system when a human being faces a stressor. Recently, however, the complete ensemble empirical mode decomposition (EMD) with adaptive noise (CEEMDAN) allows analyzing nonstationary signals in a nonlinear and time- variant way. Consequently, CEEMDAN may provide a means to obtain clues about ANS regulation in health and disease. In this study, we analyze the average Hilbert-Huang spectrum (HHS) of cardiovascular variability signals by CEEMDAN during a head-up tilt test (HUTT) in 12 healthy female subjects and 18 orthostatic intolerance female patients. Beat-to-beat intervals (BBI) as well as systolic (SYS) blood pressure variability time series were analyzed. In addition, instantaneous amplitudes and frequencies of specific intrinsic mode functions (IMF) were investigated separately to define the influence of the disease on ANS regulation. Female groups demonstrated statistical differences in the high-frequency band of BBI but higher differences for the high and low-frequency bands of SYS from the mechanical transition of HUTT.Clinical Relevance- A relevant outcome of the study is the average HHS of healthy female subjects along HUTT. This HHS may be used as reference to help diagnose OI when HHS of the cardiovascular variability signals of any subject deviates from the normal course.Over a third of patients suffering from epilepsy continue to live with recurrent disabling seizures and would greatly benefit from personalized seizure forecasting. While electroencephalography (EEG) remains most popular for studying subject-specific epileptic precursors, dysfunctions of the autonomous nervous system, notably cardiac activity measured in heart rate variability (HRV), have also been associated with epileptic seizures. This work proposes an unsupervised clustering technique which aims to automatically identify preictal HRV changes in 9 patients who underwent simultaneous electrocardiography (ECG) and intracranial EEG presurgical monitoring at the University of Montreal Hospital Center. A 2-class k-means clustering combined with a quantitative preictal HRV change detection technique were adopted in a subject- and seizure-specific manner. Results indicate inter and intra-patient variability in preictal HRV changes (between 3.5 and 6.5 min before seizure onset) and a statistically significant negative correlation between the time of change in HRV state and the duration of seizures (p less then 0.05). The presented findings show promise for new avenues of research regarding multimodal seizure prediction and unsupervised preictal time assessment.Clinical Relevance- This study proposed an unsupervised technique for quantitatively identifying preictal HRV changes which can be eventually used to implement an ECG-based seizure forecasting algorithm.In this paper, a deep learning framework for detection and classification of EMG signals for diagnosis of neuromuscular disorders is proposed employing cross wavelet transform. Cross wavelet transform which is a modification of continuous wavelet transform is an important tool to analyze any non-stationary signal in time scale and in time-frequency frame. To this end, EMG signals of healthy, myopathy and Amyotrophic lateral sclerosis disorders were procured from an online existing database. A healthy EMG signal was chosen as reference and cross wavelet transform of the rest of the healthy as well as the disease EMG signals was done with the reference. From the resulting cross wavelet spectrum images of EMG signals, a convolution neural network (CNN) based automated deep feature extraction technique was implemented. The extracted deep features were further subjected to feature ranking employing one way analysis of variance (ANOVA) test. The extracted deep features with high degree of statistical significance were fed to several benchmark machine learning classifiers for the purpose of discrimination of EMG signals. Two binary classification problems are addressed in this paper and it has been observed that the highest mean classification accuracy of 100% is achieved using the statistically significant extracted deep features. The proposed method can be implemented for real-time detection of neuromuscular disorders.The nonstationarity measure of surface Electromyography (sEMG) signals provide an index for muscle fatigue conditions. In this paper, a new framework has been proposed for the analysis of sEMG signal using Instantaneous Spectral Centroid (ISC). The novelty of the proposed work is use of topological signal processing method to quantify the nonstationarity of sEMG signal. For this, the signals are recorded from the biceps brachii muscles of 25 healthy subjects in isometric contraction. The analytical signals corresponding to nonfatigue and fatigue segments are computed using Hilbert Transform. Further, topological features such as center of gravity (CoG), triangular area function (TAF) and ISC are calculated from the geometrical representation of a transformed signal. The result indicates the increase of TAF in fatigue condition and the significant right shift of CoG in x-axis for 80% of subjects. Importantly, the ISC estimate is decreased by 17% upon fatiguing for 84% of subjects. The obtained results show statistical significance with p less then 0.05. It is observed that the shape parameters are varied in accordance with the changes observed in global characteristics of sEMG signals during muscle fatigue. The preliminary results show that the topological features are able to quantify the nonstationarity in sEMG signal. Therefore, the proposed method can be used as a fatigue index for diagnosing various neuromuscular disorders.Clinical Relevance-This method can be used to establish metrics of muscle fatigue for the benefit of physicians especially in the field of fitness, sports, pre and post-surgery surveillance and rehabilitation.This study investigates the applicability of Electromyography (EMG) signal classification algorithms with relatively low training time to control prosthetic devices. The perceived quality of control depends on many factors, such as the 1) accuracy of the algorithm, 2) the complexity of the control, and 3) the ability to compensate for the error. The high granularity of control in the time domain reduces the perceived effect of error but also limits the classification accuracy. This work aims to find the borderline for the accuracy of algorithms to be selected as a control strategy for hand prosthetic devices and thus shorten the gap between laboratory devices and commercially available devices. In particular, we compared five classification algorithms and selected one for real-time testing. The results from a test conducted on four subjects showed that the EMG-based control strategy has comparable performances with an IMU-based controller.Surface electromyography has become one of the popular methods for recognizing hand gestures. In this paper, the performance of four classification methods on sEMG signals have been investigated. These methods are developed by combinations of two feature extraction methods, including Mean Absolute Value and Short-Time Fourier Transform, and two classifiers, including Support Vector Machine and Convolutional Neural Network. These classification methods achieved an accuracy over 97 % on the NinaPro dataset 1. In addition, a new dataset, which includes the Activities of Daily Living, was proposed and an accuracy over 98 % was obtained by applying the presented classification methods.This methodology can provide the basis for a robust quantitative technique to evaluate hand grasps of stroke patients in performing activities of daily living that in turn can lead to a more efficient rehabilitation regimen.The EMG signal is very difficult to classify due to the stochastic complexity of its characteristics. A way to reduce the complexity of a signal is to use clusters to resize them to a smaller space and then perform the classification. A classification improvement was verified by clustering the electromyographic signal and comparing it with the possible movements that can be performed. In this study, the Agglomerative Hierarchical Clustering was used. The basic idea is to give prior information to the final classifier so the posterior classification has fewer classes, diminishing his complexity. Through the methodology applied in this article, an accuracy of more than 90% was achieved by using a time window of only 10 ms in a signal sampled at 2000 Hz. Experimentation confirms that the methods presented in this paper are competitive with other methods presented in the literature.Before the operation of a biosignal-based application, long-duration calibration is required to adjust the pre-trained classifier to a new user data (target data). For reducing such time-consuming step, linear domain adaptation (DA) transfer learning approaches, which transfer pooled data (source data) related to the target data, are highlighted. In the last decade, they have been applied to surface electromyogram (sEMG) data with the implicit assumption that sEMG data are linear. However, sEMGs typically have non-linear characteristics, and due to the discrepancy between the assumption and actual characteristics, linear DA approaches would cause a negative transfer. This study investigated how the correlation between the source and target data affects an 8-class forearm movement classification after applying linear DA approaches. As a result, we found significant positive correlations between the classification accuracy and the source-target correlation. Additionally, the source-target correlation depended on the motion class. Therefore, our results suggest that we should choose a non-linear DA approach when the source-target correlation among subjects or motion classes is low.A number of techniques have been reported to detect mental stress. Surface Electromyography (sEMG) has also been used to measure stress by acquiring the signals from various sites of the human body, however, consensus need to be established to determine the best possible site to harvest stress related information. In this study, work related mental stress using sEMG signals acquired from trapezius muscle and facial muscles were compared. BIOPAC signal acquisition system was used to acquire sEMG signals simultaneously from both trapezius and facial muscles from forty five (45) healthy volunteers. Stress was induced using different standard methods in a controlled environment. Statistical significant difference was found between the stress and rest levels of sEMG signals. The statistical test also showed that the upper trapezius muscle was a better stress detection site as compared to facial muscles.Clinical Relevance- Optimized stress detection can help in the prevention of the possible stress related physical disorders.This paper presents a genetic algorithm (GA) feature selection strategy for sEMG hand-arm movement prediction. The proposed approach evaluates the best feature set for each channel independently. Regularized Extreme Learning Machine was used for the classification stage. The proposed procedure was tested and analyzed applying Ninapro database 2, exercise B. Eleven time domain and two frequency domain metrics were considered in the feature population, totalizing 156 combined feature/channel. As compared to previous studies, our results are promising - 87.7% accuracy was achieved with an average of 43 combined feature/channel selection.Patients suffering from chronic facial palsy are frequently impaired by severe life-long dysfunctions. Thus, the loss of the ability to close eyes rapidly and completely bears the risk of corneal damages. Moreover, the loss of smile and an altered facial expression imply psychological stress and impede a healthy social life. Since surgical and conservative treatments frequently do not solve many problems sufficiently, closed-loop neural prosthesis are considered as feasible approach. For it, amongst others a reliable detection of the currently executed facial movement is necessary. In our proof of concept study, we propose a data-driven feature extraction for classifying eye closures and smile based on intramuscular EMGs from orbicularis oculi and zygomaticus muscles of the patient's palsy side. The data-adaptive nature of the approach enables a flexible applicability to different muscles and subjects without patient-or muscle-specific adaptations.Controlling powered prostheses with myoelectric pattern recognition (PR) provides a natural human-robot interfacing scheme for amputees who lost their limbs. Research in this direction reveals that the challenges prohibiting reliable clinical translation of myoelectric interfaces are mainly driven by the quality of the extracted features. Hence, developing accurate and reliable feature extraction techniques is of vital importance for facilitating clinical implementation of Electromyogram (EMG) PR systems. To overcome this challenge, we proposed a combination of Range Spatial Filtering (RSF) and Recurrent Fusion of Time Domain Descriptors (RFTDD) in order to improve the classifier performance and make the prosthetic hand control more appropriate for clinical applications. RSF is used to increase the number of EMG signals available for feature extraction by focusing on the spatial information between all possible logical combinations of the physical EMG channels. RFTDD is then used to capture the temporal inforlow-cost clinical applications.This work demonstrates the effectiveness of Convolutional Neural Networks in the task of pose estimation from Electromyographical (EMG) data. The Ninapro DB5 dataset was used to train the model to predict the hand pose from EMG data. The models predict the hand pose with an error rate of 4.6% for the EMG model, and 3.6% when accelerometry data is included. This shows that hand pose can be effectively estimated from EMG data, which can be enhanced with accelerometry data.Recently, the subject-specific surface electromyography (sEMG)-based gesture classification with deep learning algorithms has been widely researched. However, it is not practical to obtain the training data by requiring a user to perform hand gestures many times in real life. This problem can be alleviated to a certain extent if sEMG from many other subjects could be used to train the classifier. In this paper, we propose a normalisation approach that allows implementing real-time subject-independent sEMG based hand gesture classification without training the deep learning algorithm subject specifically. We hypothesed that the amplitude ranges of sEMG across channels between forearm muscle contractions for a hand gesture recorded in the same condition do not vary significantly within each individual. Therefore, the min-max normalisation is applied to source domain data but the new maximum and minimum values of each channel used to restrict the amplitude range are calculated from a trial cycle of a new user (target domain) and assigned by the class label. A convolutional neural network (ConvNet) trained with the normalised data achieved an average 87.03% accuracy on our G. dataset (12 gestures) and 94.53% on M. dataset (7 gestures) by using the leave-one-subject-out cross-validation.When generating automatic sleep reports with mobile sleep monitoring devices, it is crucial to have a good grasp of the reliability of the result. In this paper, we feed features derived from the output of a sleep scoring algorithm to a 'regression ensemble' to estimate the quality of the automatic sleep scoring. We compare this estimate to the actual quality, calculated using a manual scoring of a concurrent polysomnography recording. We find that it is generally possible to estimate the quality of a sleep scoring, but with some uncertainty ('root mean squared error' between estimated and true Cohen's kappa is 0.078). We expect that this method could be useful in situations with many scored nights from the same subject, where an overall picture of scoring quality is needed, but where uncertainty on single nights is less of an issue.Deep learning has become popular for automatic sleep stage scoring due to its capability to extract useful features from raw signals. Most of the existing models, however, have been overengineered to consist of many layers or have introduced additional steps in the processing pipeline, such as converting signals to spectrogram-based images. They require to be trained on a large dataset to prevent the overfitting problem (but most of the sleep datasets contain a limited amount of class-imbalanced data) and are difficult to be applied (as there are many hyperparameters to be configured in the pipeline). In this paper, we propose an efficient deep learning model, named TinySleepNet, and a novel technique to effectively train the model end-to-end for automatic sleep stage scoring based on raw single-channel EEG. Our model consists of a less number of model parameters to be trained compared to the existing ones, requiring a less amount of training data and computational resources. Our training technique incorporates data augmentation that can make our model be more robust the shift along the time axis, and can prevent the model from remembering the sequence of sleep stages. We evaluated our model on seven public sleep datasets that have different characteristics in terms of scoring criteria and recording channels and environments. The results show that, with the same model architecture and the training parameters, our method achieves a similar (or better) performance compared to the state-of-the-art methods on all datasets. This demonstrates that our method can generalize well to the largest number of different datasets.Feature extraction from ECG-derived heart rate variability signal has shown to be useful in classifying sleep apnea. In earlier works, time-domain features, frequency-domain features, and a combination of the two have been used with classifiers such as logistic regression and support vector machines. However, more recently, deep learning techniques have outperformed these conventional feature engineering and classification techniques in various applications. This work explores the use of convolutional neural networks (CNN) for detecting sleep apnea segments. CNN is an image classification technique that has shown robust performance in various signal classification applications. In this work, we use it to classify one-dimensional heart rate variability signal, thereby utilizing a one-dimensional CNN (1-D CNN). The proposed technique resizes the raw heart rate variability data to a common dimension using cubic interpolation and uses it as a direct input to the 1-D CNN, without the need for feature extraction and selection. The performance of the method is evaluated on a dataset of 70 overnight ECG recordings, with 35 recordings used for training the model and 35 for testing. The proposed method achieves an accuracy of 88.23% (AUC=0.9453) in detecting sleep apnea epochs, outperforming several baseline techniques.In this study, we use the overnight blood oxygen saturation (SpO2) signal along with convolutional neural networks (CNN) for the automatic estimation of pediatric sleep apnea-hypopnea syndrome (SAHS) severity. The few preceding studies have focused on the application of conventional feature extraction methods to obtain information from the SpO2 signal, which may omit relevant data related to the illness. In contrast, deep learning techniques are able to automatically learn features from raw input signal. Thus, we propose to assess whether CNN, a deep learning algorithm, could automatically estimate the apnea-hypopnea index (AHÍ) from nocturnal oximetry to help establish pediatric SAHS presence and severity. A database of 746 SpO2 recordings is involved in the study. CNN was trained using 20-min segments from the SpO2 signal in the training set (400 subjects). Hyperparameters of the CNN architecture were tuned using a validation set (100 subjects). This model was applied to a test set (246 subjects), in which the final AHI of each patient was obtained as the average of the output of the CNN for all the segments of the corresponding SpO2 signal. The AHI estimated by the CNN showed a promising diagnostic performance, with 74.8%, 90.7%, and 95.1% accuracies for the common AHI severity thresholds of 1, 5, and 10 events per hour (e/h), respectively. Furthermore, this model reached 28.6, 32.9, and 120.0 positive likelihood ratios for the above-mentioned AHI thresholds. This suggests that the information extracted from the oximetry signal by deep learning techniques may be useful to both establish pediatric SAHS and its severity.Studying the neural correlates of sleep can lead to revelations in our understanding of sleep and its interplay with different neurological disorders. Sleep research relies on manual annotation of sleep stages based on rules developed for healthy adults. Automating sleep stage annotation can expedite sleep research and enable us to better understand atypical sleep patterns. Our goal was to create a fully unsupervised approach to label sleep and wake states in human electro-corticography (ECoG) data from epilepsy patients. Here, we demonstrate that with continuous data from a single ECoG electrode, hidden semi-Markov models (HSMM) perform best in classifying sleep/wake states without excessive transitions, with a mean accuracy (n=4) of 85.2% compared to using K-means clustering (72.2%) and hidden Markov models (81.5%). Our results confirm that HSMMs produce meaningful labels for ECoG data and establish the groundwork to apply this model to cluster sleep stages and potentially other behavioral states.In this paper, we propose a novel method of automatic sleep stage classification based on single-channel electroencephalography (EEG). First, we use marginal Hilbert spectrum (MHS) to depict time-frequency domain features of five sleep stages of 30-second (30s) EEG epochs. Second, the extracted MHSs features are input to a convolutional neural network (CNN) as multi-channel sequences for the sleep stage classification task. Third, a focal loss function is introduced into the CNN classifier to alleviate the classes imbalance problem of sleep data. Experimental results show that the proposed method can obtain an overall accuracy of 86.14% on the public Sleep-EDF dataset, which is competitive and worth exploring among a series of deep learning methods for the automatic sleep stage classification task.The use of fetal heart rate (FHR) recordings for assessing fetal wellbeing is an integral component of obstetric care. Recently, non-invasive fetal electrocardiography (NI-FECG) has demonstrated utility for accurately diagnosing fetal arrhythmias via clinician interpretation. In this work, we introduce the use of data-driven entropy profiling to automatically detect fetal arrhythmias in short length FHR recordings obtained via NI-FECG. Using an open access dataset of 11 normal and 11 arrhythmic fetuses, our method (TotalSampEn) achieves excellent classification performance (AUC = 0.98) for detecting fetal arrhythmias in a short time window (i.e. under 10 minutes). We demonstrate that our method outperforms SampEn (AUC = 0.72) and FuzzyEn (AUC = 0.74) based estimates, proving its effectiveness for this task. The rapid detection provided by our approach may enable efficient triage of concerning FHR recordings for clinician review.Despite the enormous potential applications, non-invasive recordings have not yet made enough satisfaction for fetal disease detection. This is mainly due to the fetal ECG signal is contaminated by the maternal electrocardiograph (ECG) interference, muscle contractions, and motion artifacts. In this paper, we propose a joint multiple subspace-based blind source separation (BSS) approach to extract the fetal heart rate (HR), so that it could greatly reduce the effect of maternal ECG and motion artifacts. The approach relies on the estimation of the coefficient matrix formulated as the tensor decomposition in terms of multiple datasets. Since the objective function takes the coupling information from the stacking of the covariance matrix for multiple datasets into account, estimating the coefficient matrices is fulfilled not only on dependence across multiple datasets, but also can combine the extracted components across four different datasets. Numerical results demonstrate that the proposed method can achieve a high extracted HR accuracy for each dataset, when compared to some conventional methods.In the fetal period, the progressive coordination among several subsystems promotes the emergence of sleep states. For this reason, the characterization of fetal behavioral states plays a crucial role in assessing fetal wellbeing. Nevertheless, current methodologies aimed at assessing fetal sleep states over limited time intervals require visual observation of the traces. In this work, we validate a point process approach for a continuous in time characterization of fetal behavioral states. We compare traditional heart rate variability (HRV) parameters and the corresponding point process-extracted sets of time and frequency measures in a population of 39 fetuses whose fetal ECG was recorded overnight during the third trimester of gestation.Clinical Relevance- Our results provide evidence for the proposed point process framework to capture fetal HRV dynamics with a high degree of reliability, suggesting its potential application for instantaneous estimates of fetal sleep states.Beat-by-beat maternal and fetal heart couplings were reported to be evident throughout the fetal development. However, it is still unknown whether maternal-fetal heartbeat coupling parameters are associated with fetal development, and the potential interrelationships. Therefore, this study aims to investigate the associations of coupling parameters with fetal gestational age by multivariate regression models. Ten min abdominal lead-based maternal and fetal ECG signals were collected from 16 healthy pregnant women with healthy singleton pregnancies (19-32 weeks). Maternal and Fetal Heart Rate Variability (MHRV and FHRV) values as well as maternal-fetal heart rate coupling (strength, measured by A) parameters at various coupling ratios (associated with different MaternalFetal heartbeat ratios of 12, 13, 23, 24, 34, and 35) were calculated. Based on those features stepwise multivariate regression models were constructed by validating against the gold standard gestational age identified by crown-rump length from doppler echocardiogram. Among all models, the best model (Root Mean Square Error, RMSE=1.92) was found to be significantly (p less then 0.05) associated with mean fetal heart rate, mean maternal heart rate, standard deviation of maternal heart rate, λ[13], λ[23], λ[24]. Correlation coefficients and Bland Altman plots were constructed to statistically validate the results. The model developed based on coupling parameters only, showed the second-best performance (RMSE=2.50). Therefore, combining maternal and fetal heart rate variability parameters with maternal-fetal heart rate coupling values (rather than considering FHRV or MHRV parameters only) is found to be better associated with fetal development.Clinical relevance- This is a brief additional statement on why this might be of interest to practicing clinicians. Example This establishes the anesthetic efficacy of 10% intraosseous injections with epinephrine to positively influence cardiovascular function.Nearly 10% of all births in the United States are preterm. Preterm birth is a major risk for developmental neuromotor disorders. Early characterization of a future developmental outcome is necessary to design early interventions. However, such evaluations are currently subjective and typically happen only several months after birth. The aim of this study was to quantify movement bouts after birth and to determine if features of maturation might be characterized. Four preterm infants were continuously monitored for several months, from a few days after birth until discharge, in the Neonatal Intensive Care Unit. Movement was quantified from the photoplethysmogram using a wavelet-based algorithm. In all 4 infants, maturation was associated with a decrease (p 30s). The distribution of movement durations followed a power law function with its exponent defining the characteristic of the distribution. The exponent significantly increased with post-menstrual age. Future research will test whether these maturational changes can predict developmental outcomes.Clinical Relevance- Early identification of changes in features of preterm infant movement may be useful in predicting neuromotor development and potential disorders.A pilot study on tracking changes in tidal volume (TV) using ECG signals acquired by a wearable armband is presented. The wearable armband provides three ECG channels by using three pairs of dry electrodes, resulting in a device that is convenient for long-term daily monitoring. An additional ECG channel was derived by computing the first principal component of the three original channels (by means of principal component analysis). Armband and spirometer signals were simultaneously recorded from five healthy subjects who were instructed to breathe with varying TV. Three electrocardiogram derived respiration (EDR) methods based on QRS complex morphology were studied the QRS slopes range (SR), the R-wave angle (Փ), and the R-S amplitude (RS). The peak-to-peak amplitudes of these EDR signals were estimated as surrogates for TV, and their correlations with the reference TV (estimated from the spirometer signal) were computed. In addition, a multiple linear regression model was calculated for each subject, using the peak-to-peak amplitudes from the three EDR methods from the four ECG channels. Obtained correlations between TV and EDR peak-to-peak amplitude ranged from 0.0448 up to 0.8491. For every subject, a moderate correlation (>0.5) was obtained for at least one EDR method. Furthermore, the correlations obtained for the subject-specific multiple linear regression model ranged from 0.8234 up to 0.9154, and the goodness of fit was 0.73±0.07 (median ± standard deviation). These results suggest that the peak-to-peak amplitudes of the EDR methods are linearly related to the TV. opening the possibility of estimating TV directly from an armband ECG device.Clinical Relevance- This opens the door to possible continuous monitoring of TV from the armband by using EDR.We propose a novel electrocardiogram (ECG) denoising technique using the variable frequency complex demodulation (VFCDM) algorithm. We used VFCDM to perform the sub-band decomposition of the noise-contaminated ECG to remove the noise components so that accurate QRS complexes could be identified. The ECG quality was further improved by removing baseline drift and smoothing via adaptive mean filtering. The proposed method was validated on the MIT-BIH arrhythmia database (MITDB) and wearable armband ECG data. For the former, we added Gaussian white noise to the ECG signals at different signal-to-noise ratios and the denoising performance of the proposed method was compared with other denoising techniques. The proposed approach showed superior denoising performance compared to the other methods. We compared the QRS complex detection performance of the noisy to the denoised armband ECG. The performance of the proposed denoising method on the armband ECG resulted in comparable QRS complex detection as that obtained when using Holter monitor ECG signals. This demonstrates that the proposed algorithm can significantly increase the amount of usable armband ECG data, which would otherwise have been discarded due to electromyogram contamination especially during arm movements. Hence, the proposed algorithm has the potential to enable long-term monitoring of atrial fibrillation using the armband without the discomfort of skin irritation often experienced with Holter monitors.Clinical Relevance- The proposed ECG denoising method can significantly increase the ECG quality of wearable ECG devices, which are more susceptible to muscle and motion artifacts.Stroke survivors are often characterized by hemiparesis, i.e., paralysis in one half of the body, that severely affects upper limb movements. Continuous monitoring of the progression of hemiparesis requires manual observation of the limb movements at regular intervals and hence is a labour intensive process. In this work, we use wrist-worn accelerometers for automated assessment of hemiparetic severity in acute stroke patients through bivariate Poincaré analysis between accelerometer data from the two hands during spontaneous and instructed movements. Experiments show that while the bivariate Poincaré descriptors CSD1 and CSD2 can identify hemiparetic patients from control subjects, a novel descriptor called Complex Cross-Correlation Measure (C3M) can distinguish between moderate and severe hemiparesis. Further, we justify the use of C3M by showing that it is described by multiple-lag cross-correlations, representing the co-ordination of activity between two hands. The descriptors are compared against the National Institutes of Health Stroke Scale (NIHSS), the clinical gold standard for evaluation of hemiparetic severity, and studied using statistical tests for developing supervised models for hemiparesis classification.
Read More: