Waveform Segments Research Articles

Automated ECG Waveform Annotation Based on Stacked Long Short-Term Memory

The classification of electrocardiogram (ECG) waveform segmentation techniques can be difficult due to physiological variation of heart rate and different characteristics of the different ECG waves in terms of shape, frequency, amplitude, and duration. The P-wave, PR-segment, QRS-complex, ST-segment, and T-wave are extracted as the feature for classification algorithm to diagnose specified cardiac disorders. This requires the implementation of algorithms that identify specific points within the ECG wave. Some previous computational algorithms for automatic classification of ECG segmentation are proposed to overcome limitations of manual inspection of the ECG. This study presents new insight into the ECG semantic segmentation problem is surmounted by a deep learning approach for automatic ECG wave-form. Long short-term memory (LSTM) is proposed for this task. This experimental study has been performed for six different waveforms of ECG signal that represents cardiac disorders obtained from the Physionet: QT database. Overall, LSTM performance achieved accuracy, sensitivity, specificity, precision, F1-score, is 93.36%, 86.85%, 95.78%, 81.79%, and 83.09%, respectively.

Automated ECG Waveform Annotation Based on Stacked Long Short-Term Memory

The classification of electrocardiogram (ECG) waveform segmentation techniques can be difficult due to physiological variation of heart rate and different characteristics of the different ECG waves in terms of shape, frequency, amplitude, and duration. The P-wave, PR-segment, QRS-complex, ST-segment, and T-wave are extracted as the feature for classification algorithm to diagnose specified cardiac disorders. This requires the implementation of algorithms that identify specific points within the ECG wave. Some previous computational algorithms for automatic classification of ECG segmentation are proposed to overcome limitations of manual inspection of the ECG. This study presents new insight into the ECG semantic segmentation problem is surmounted by a deep learning approach for automatic ECG wave-form. Long short-term memory (LSTM) is proposed for this task. This experimental study has been performed for six different waveforms of ECG signal that represents cardiac disorders obtained from the Physionet: QT database. Overall, LSTM performance achieved accuracy, sensitivity, specificity, precision, F1-score, is 93.36%, 86.85%, 95.78%, 81.79%, and 83.09%, respectively.

Using A Comparison between Low-Carb and High-Carb Meals with A Detailed PPG Waveform Segmentation and Associated Energy Analysis to Investigate the Linkage between the Brain and Internal Organs (GH-Method: Math-Physical Medicine)

In this paper, the author analyzed his 1,707 meals that are segregated into low-carb and high-carb meals. The purpose is to attempt the validation of his hypothesis on the relationship from the directive of the brain to the gastrointestinal functions and liver’s glucose production capability by using PPG value changes over a smaller time segment (each segment with 15 minutes). The author uses the math-physical approach rather than a biochemical one

Enhanced Automated Diagnosis of Coronary Artery Disease Using Features Extracted From QT Interval Time Series and ST–T Waveform

There is a growing interest in automated diagnosis of coronary artery disease (CAD) with the application of machine learning (ML) methods to the body surface electrocardiograph (ECG). Although prior studies have documented associations of CAD with increased QT variability and ST–T segment abnormalities such as T-wave inversion and ST-segment elevation or depression, their efficacy in automated CAD detection has not been fully investigated. To validate their usefulness, a dataset containing related clinical characteristics and 5-min single-lead ECGs of 107 healthy controls and 93 CAD patients was first constructed. Based on this dataset, simultaneous analyses were then conducted in five scenarios, in which different ML algorithms were applied to classify the two groups with various features derived from the RR and QT interval time-series and ST–T segment waveforms. Compared with utilizing features obtained from the RR interval time-series, better classification results were achieved utilizing that obtained from the QT interval time-series. The classification results were elevated with combining utilization of features derived from both the RR and QT interval time-series. By further fusing features extracted from ST–T segment waveforms, the best performance was achieved with 96.16% accuracy, 95.75% sensitivity, and 96.40% specificity. Based the best performance, an automated CAD detection system was developed with extreme gradient boosting, an ensemble ML algorithm, and the residual neural network, namely, a deep learning method. The results of this study support the potential of information derived from the QT interval time-series and ST–T segment waveforms in ECG-based automated CAD detection.

Multimodal Word Discovery and Retrieval With Spoken Descriptions and Visual Concepts

In the absence of dictionaries, translators, or grammars, it is still possible to learn some of the words of a new language by listening to spoken descriptions of images. If several images, each containing a particular visually salient object, each co-occur with a particular sequence of speech sounds, we can infer that those speech sounds are a word whose definition is the visible object. A multimodal word discovery system accepts, as input, a database of spoken descriptions of images (or a set of corresponding phone transcriptions) and learns a mapping from waveform segments (or phone strings) to their associated image concepts. In this article, four multimodal word discovery systems are demonstrated: three models based on statistical machine translation (SMT) and one based on neural machine translation (NMT). The systems are trained with phonetic transcriptions, MFCC and multilingual bottleneck features (MBN). On the phone-level, the SMT outperforms the NMT model, achieving a 61.6% F1 score in the phone-level word discovery task on Flickr30k. On the audio-level, we compared our models with the existing ES-KMeans algorithm for word discovery and present some of the challenges in multimodal spoken word discovery.

Automatic classification of regular and irregular capnogram segments using time- and frequency-domain features: A machine learning-based approach.

The quantitative features of a capnogram signal are important clinical metrics in assessing pulmonary function. However, these features should be quantified from the regular (artefact-free) segments of the capnogram waveform. This paper presents a machine learning-based approach for the automatic classification of regular and irregular capnogram segments. Herein, we proposed four time- and two frequency-domain features experimented with the support vector machine classifier through ten-fold cross-validation. MATLAB simulation was conducted on 100 regular and 100 irregular 15 s capnogram segments. Analysis of variance was performed to investigate the significance of the proposed features. Pearson's correlation was utilized to select the relatively most substantial ones, namely variance and the area under normalized magnitude spectrum. Classification performance, using these features, was evaluated against two feature sets in which either time- or frequency-domain features only were employed. Results showed a classification accuracy of 86.5%, which outperformed the other cases by an average of 5.5%. The achieved specificity, sensitivity, and precision were 84%, 89% and 86.51%, respectively. The average execution time for feature extraction and classification per segment is only 36 ms. The proposed approach can be integrated with capnography devices for real-time capnogram-based respiratory assessment. However, further research is recommended to enhance the classification performance.

Noninvasive Classification of Blood Pressure Based on Photoplethysmography Signals Using Bidirectional Long Short-Term Memory and Time-Frequency Analysis

The photoplethysmography (PPG) method for continuous noninvasive measurements of blood pressure (BP) offers a more comfortable solution than conventional methods. The main challenge in using the PPG method is that its accuracy is greatly influenced by motion artifacts. In addition, the characteristics of PPG vary depending on physiological conditions; hence, the system must be calibrated to adjust for such changes. We attempt to address these limitations and propose a novel method for the classification of BP using a bidirectional long short-term memory (BLSTM) network with time-frequency (TF) analysis based on PPG signals. The TF analysis extracts information from PPG signals using a short-time Fourier transform (STFT) in the time domain to produce two features, namely, the instantaneous frequency and spectral entropy. Training the BLSTM network using TF features significantly improves the classification performance and decreases the training time. We classify 900 PPG waveform segment samples from 219 adult subjects into three classification levels: normotension (NT), prehypertension (PHT) and hypertension (HT). The results show that the proposed method is successful in the classification of BP with accuracy, sensitivity, and speciticity values of 97.33%, 100%, and 94.87%, respectively. The F1 scores of three BP classifications were 97.29%, 97.39%, and 93.93%, respectively. A comparison of current and previous approaches to the classification of BP is accomplished. Our proposed method achieves a higher accuracy than convolutional neural networks (CNNs), k-nearest neighbors (KNN), bagged tree, logistic regression, and AdaBoost tree methods.

Assessment of artifacts reduction and denoising techniques in Electrocardiographic signals using Ensemble Average-based method

Background and objective: Outpatient vital signs monitoring has a key role in medical diagnosis and treatment. However, ambulatory vital signs monitoring has great challenges to overcome, being the most important, the reduction of noise and Motion Artifacts, which hide essential information, particularly in Electrocardiographic signals. Despite efforts being made to reduce these artifacts, a comparative performance assessment of proposed techniques does not exist to the best of our knowledge and there are no enhancement level measurements obtained by the signals in the artifacts reduction. This article presents a new method based on Ensemble Average for the performance comparison of reported techniques for the processing and reduction of noise and artifacts in Electrocardiographic signals.Methods: The comparison was done using a dataset composed by six synthetic noised Electrocardiographic signals and six real one acquired from healthy volunteers that intentionally introduced Motion Artifacts. Several techniques that have reported positive results in the enhancement of Electrocardiographic signals were applied to this dataset to compare their performance in the reduction of Motion Artifacts. The Signal-to-Noise Ratio and the Ensemble Average as a distortion measurement were used to compare the performance of algorithms to produce an enhanced signal.Results: In agreement to previous reports, all studied methods show a significant improvement of the Signal-to-Noise Ratio. Concerning the distortion of the waveform, although all methods caused high distortion on the enhanced signal waveform, the Wavelet-ICA method showed the best performance. The percentage of signal distortion introduced by denoising techniques was evaluated through the proposed Ensemble Average Electrocardiographic method.Conclusions: It was found that the proposed method based on Ensemble Average offers a complementary way to measure the performance of denoising techniques when considering the introduced distortion in the waveform segments once the artifact reduction process was applied and not only the change in the Signal-to-Noise Ratio.

1D Convolutional Neural Networks for Estimation of Compensatory Reserve from Blood Pressure Waveforms.

We propose a Deep Convolutional Neural Network (CNN) architecture for computing a Compensatory Reserve Metric (CRM) for trauma victims suffering from hypovolemia (decreased circulating blood volume). The CRM is a single health indicator value that ranges from 100% for healthy individuals, down to 0% at hemodynamic decompensation - when the body can no longer compensate for blood loss. The CNN is trained on 20 second blood pressure waveform segments obtained from a finger-cuff monitor of 194 subjects. The model accurately predicts CRM when tested on data from 22 additional human subjects obtained from Lower Body Negative Pressure (LBNP) emulation of hemorrhage, attaining a mean squared error (MSE) of 0.0238 over the full range of values, including those from subjects with both low and high tolerance to central hypovolemia.

Improving accuracy of derived 12-lead electrocardiography by waveform segmentation

A number of methods have been proposed to reduce number of leads for electrocardiography (ECG) measurement without decreasing the signal quality. Some limited sets of leads that are nearly orthogonal, such as EASI, have been used to reconstruct the standard 12-lead ECG by various transformation techniques including linear, nonlinear, generic, and patient-specific. Those existing techniques, however, employed a full-cycle ECG waveform to calculate the transformation coefficients. Instead of calculating the transformation coefficients using a full-cycle waveform, we propose a new approach that segments the waveform into three segments: PR, QRS complex, and ST, hence the transformation coefficients were segment-specific. For testing, our new segment-specific approach was applied to six existing methods: Dower’s method with generic coefficients, Dower’s method with individual (patient-specific) coefficients, Linear Regression (LR), 2nd degree Polynomial Regression (PR), 3rd degree PR, and Artificial Neural Network (ANN). The results showed that the new approach outperformed the conventional full-cycle approach. It was able to significantly reduce the derivation error up to 74.50% as well as improve the correlation coefficient up to 0.66%.

Improving accuracy of derived 12-lead electrocardiography by waveform segmentation

A number of methods have been proposed to reduce number of leads for electrocardiography (ECG) measurement without decreasing the signal quality. Some limited sets of leads that are nearly orthogonal, such as EASI, have been used to reconstruct the standard 12-lead ECG by various transformation techniques including linear, nonlinear, generic, and patient-specific. Those existing techniques, however, employed a full-cycle ECG waveform to calculate the transformation coefficients. Instead of calculating the transformation coefficients using a full-cycle waveform, we propose a new approach that segments the waveform into three segments: PR, QRS complex, and ST, hence the transformation coefficients were segment-specific. For testing, our new segment-specific approach was applied to six existing methods: Dower’s method with generic coefficients, Dower’s method with individual (patient-specific) coefficients, Linear Regression (LR), 2nd degree Polynomial Regression (PR), 3rd degree PR, and Artificial Neural Network (ANN). The results showed that the new approach outperformed the conventional full-cycle approach. It was able to significantly reduce the derivation error up to 74.50% as well as improve the correlation coefficient up to 0.66%.

Enhancement of Long-Term External–Internal Correlation by Phase-Shift Detection and Correction Based on Concurrent External Bellows and Internal Navigator Signals

PurposeThe purpose of this study was to enhance the correlation between external and internal respiratory motions by dynamically determining and correcting the patient-specific phase shift between external and internal respiratory waveforms acquired concurrently during respiratory-correlated 4-dimensional magnetic resonance imaging scans. Methods and MaterialsInternal-navigator and external-bellows waveforms were acquired simultaneously during 6- to 15-minute respiratory-correlated 4-dimensional magnetic resonance imaging scans in 10 healthy participants under an institutional review board–approved protocol. The navigator was placed at the right lung–diaphragm interface, and the bellows were placed ∼5 cm inferior to the sternum. Three segments of each respiratory waveform, at the beginning, middle, and end of a scan, were analyzed. Three phase-domain methods were employed to estimate the phase shift, including analytical signal analysis, phase-space oval fitting, and principal component analysis. A robust strategy for estimating the phase shift was realized by combining these methods in a weighted average and by eliminating outliers (>2 σ) caused by breathing irregularities. Whether phase-shift correction affects the external-internal correlation was evaluated. The cross-correlation between the 2 waveforms in the time domain provided an independent check of the correlation enhancement. ResultsPhase-shift correction significantly enhanced the external-internal correlation in all participants across the entire 6- to 15-minute scans. On average, the correlation increased from 0.45 ± 0.28 to 0.85 ± 0.15 for the combined method. The combined method exhibited a 99.5% success rate and revealed that the phase of the external waveform leads that of the internal waveform in all 10 participants by 57 o ± 17o (1.6 ± 0.5 bins) on average. Seven participants exhibited highly reproducible phase shifts over time, evidenced by standard deviations (σ) < 4o, whereas 8o < σ < 12o in the remaining 3 participants. Regardless, phase-shift correction significantly improved the correlation in all participants. ConclusionsCorrecting the phase shift estimated by the phase-domain methods provides a new approach for enhancing the correlation between external and internal respiratory motions. This strategy holds promise for improving the accuracy of respiratory-gated radiation therapy.

Abstract 127: The Effect of Cutoff Frequency Range on Amplitude Spectrum Area Performance in Predicting Defibrillation Outcome

Introduction: Both clinical and experimental studies have demonstrated that ventricular fibrillation (VF) waveform-based amplitude spectrum area (AMSA) predicts defibrillation outcome with high sensitivity and specificity. A commonly used cutoff frequency range for calculation of AMSA is 4-48 Hz, but the median VF frequency in patients is less than 7 Hz. In the present study, we investigated whether the AMSA performance in predicting defibrillation outcome may be affected by different cutoff frequency ranges. Hypothesis: A selected cutoff frequency range would affect AMSA performance in predicting defibrillation outcome in patients. Methods: A total of 450 segments (179-defibrillation success and 271-defibrillation failure) of VF waveforms from patients with out-of-hospital cardiac arrest were analyzed. Each 4.096-s segment prior to defibrillation was acquired with a sampling rate of 250 Hz. Using MATLAB (MathWorks, Natck, MA), the AMSA of each segment of VF waveform was calculated with five cutoff frequency ranges (Shown in Figure 1). Sensitivity and specificity were calculated when defibrillation success rate was 50% which was estimated by a logistic regression curve. Area under the receiver operating characteristic curve (AUC) was used to evaluate AMSA performance within different cutoff frequency ranges. Results: AMSA threshold was decreased from 20.5 to 6.5 mV·Hz when the cutoff frequency range was changed from 3-48 Hz to 3-10 Hz (Figure 1). Sensitivity of AMSA in 3-10 Hz was increased by 81.3% compared with that of AMSA in 3-48 Hz. Specificity of AMSA in 3-10 Hz was decreased by 8.1% compared with that of AMSA in 3-48 Hz. The corresponding AUC was increased from 0.724 (3-48 Hz) to 0.768 (3-10 Hz). Conclusions: A cutoff frequency range of 3-10 Hz significantly increases the sensitivity of AMSA for predicting defibrillation success in patients with cardiac arrest with a slight decrease in specificity.

Vocal Fold Vibration in Older Adults With and Without Age-Related Dysphonia

The purpose of this study was to identify the extent to which 7 measures of glottal area timing and regularity differ between older adults with and without age-related dysphonia (ARD). Laryngeal high-speed videoendoscopy was completed at 4,000 frames per second for 42 adults aged 70 years and older (ARD: 9 female, 5 male; control group: 15 female, 13 male). Relative glottal gap, open quotient, speed index, maximum area declination rate, harmonics-to-noise ratio, harmonic richness factor, and standard deviation of fundamental frequency were measured from a 0.5-s segment of the glottal area waveform. Eta squared (η2) was computed to estimate group effect. Small effect sizes (η2 = .18-.35) were present for relative glottal gap, open quotient, maximum area declination rate, harmonic richness factor, and standard deviation of fundamental frequency. Speed index and glottal harmonics-to-noise ratio did not explain group membership (η2 = .001 and .05, respectively). These findings provide evidence that vocal fold vibration in ARD is different than in normal aging, whereas the overlap in values for every measure is consistent with the concept that normal aging and ARD exist as a continuum of health and disease.

Noseleaf motions impart dynamic signatures on bat biosonar pulses

Old-World leaf-nosed bats (Hipposideridae) are echolocating bats with nasal biosonar pulse emission. The nostrils in these animals are surrounded by baffle shapes (“noseleaves”) that have been shown to function as beamforming devices. In addition to their elaborate static geometries, the noseleaves can also undergo non-rigid shape changes that coincide with biosonar pulse emission. Prior work by the authors based on biomimetic baffle shapes has demonstrated that baffle motions similar to the ones seen in bat noseleaves can result in time-variant device properties, i.e., distributions of signal power over direction and frequency that also change with time. However, it remained to be established whether similar effects occur in bats. To answer this question, biosonar pulse emission in hipposiderid bats (Hipposideros armiger and Hipposideros pratti) has been studied with synchronized arrays of high-speed video cameras and ultrasonic microphones. Noseleaf motion was characterized by tracking the position of nine landmark points on the noseleaf. Temporal variability in the distribution of signal energy across spatial directions and ultrasonic frequencies was quantified using pair-wise measures of similarity applied to time-windowed segments of the pulse waveforms. Pulses that were accompanied by unequivocal noseleaf motions showed a significantly larger variability across time than those that were not.

Acquisition of Seismic, Hydroacoustic, and Infrasonic Data with Hadoop and Accumulo

ABSTRACT Recent developments in distributed data systems now allow faster ingestion and processing of larger quantities of time‐series data than available in current seismic, hydroacoustic, and infrasonic analysis platforms. However, the data model and storage architecture of these systems are significantly different than those in use today. We developed a data acquisition and signal analysis platform using a relatively inexpensive cluster of commodity computing hardware running a Hadoop Distributed File System, an Accumulo database infrastructure, and the Zeppelin web‐based analytics tool suite. The Accumulo data model allows individual waveform samples and their associated metadata to be stored as discrete rows in the database. This is a significant departure from traditional storage practices, in which continuous waveform segments are stored with their associated metadata as a single entity. Our design allows for rapid table scans of large data archives within the Accumulo database for locating, retrieving, and analyzing specific waveform segments directly. The system infrastructure is horizontally scalable, which allows additional data acquisition and processing resources to be added simply through the addition of new data nodes. Easy scalability permits the system to accommodate the ingestion and analysis of new data as the network of seismic, hydroacoustic, and infrasonic sensors grows. The barrier of entry for establishing a functional data acquisition system is relatively low for which a proof of concept prototype can be developed on a limited budget and later scaled as storage and processing requirements dictate.

Reliable Detection of Implicit Waveform-Specific Learning in Continuous Tracking Task Paradigm

Implicit waveform-specific (IWS) learning during a typical continuous tracking task paradigm has been reported for decades, as evidenced by better tracking improvement on the repeated segment of a specific target waveform than random segments. However, the occurrence of the IWS learning in such a task paradigm has been challenged by several unsuccessful results in recent literature. This research concerns reliable detection of the induced IWS learning and to this end, proposes to use the similarity between the cursor and the target along the direction corresponding to the waveform pattern as the performance measure. A 3-day experiment designed with full examination on IWS learning including a practice phase, an immediate test phase and a consolidation test phase after 24 hours was conducted to validate the feasibility and sensitivity of the Pearson’s correlation coefficient on the vertical movement rv in this study. Experiment results indicate that rv is more sensitive in detecting the IWS learning in all phases compared to the conventional root mean square error (RMSE) performance measure. The findings confirm the importance of the performance measure in implicit learning research and the similarity measure in accordance with the waveform could be promising for waveform-specific learning detection in this paradigm.

USE OF AMBIENT VIBRATIONS IN UNDERSTANDING LOCAL SITE EFFECTS AT BROADBAND SEISMIC STATIONS OF THE HELLENIC UNIFIED SEISMOLOGICAL NETWORK (HUSN)

The evaluation of ambient seismic noise at the Hellenic Unified Seismic Network (HUSN) stations is investigated in this study. Ambient vibration recordings combined with the horizontal to vertical (H/V) spectral ratio technique helps in characterizing local site effects. This technique was applied at 17 sites ambient noise measurements. We selected a number of 1-hr waveform segments during day and night for summer and winter. For each site the H/V spectral ratio was calculated and the results were combined with geological and geophysical information. The goal was to show the network performance as far as the station quality and noise level at each site concerns in order to provide possible structural improvements, seismic station relocations or to detecting operational problems.

Comparative analysis of bag‐of‐words models for ECG‐based biometrics

The performances of the bag-of-words approach in biometric applications using electrocardiography (ECG) signals have been analysed according to the influence of specific design parameters. Optimal setup scenarios have been identified combining five encoding procedures, two pooling methods, and three classification strategies. The method does not require waveform segmentation nor fiducial points detection. Comparative results based on extensive experiments conducted on real ECG recordings collected on chest, finger, and hand palm are presented. Sparse representations yield best results, exceeding 99% correct classification rate for a number of 100 subjects, while additionally exhibiting robustness against modifications of the experimental setup.

Cardiac safety profile for Random Complex Waveforms.

A rigorous method for assessing the Ventricular Fibrillation (VF) risk of a Random Complex Waveform (RCW) has not been previously available. Real-life hazardous events motivated us to develop such method. An RCW is observable and recordable. It consists of multiple different components randomly added one to the other. Assessment for VF risk exists for non-random waveforms, particularly VF thresholds for 50/60 Hz alternating currents, but not for RCWs. We developed a method which considers exposure to a segment of an RCW. It transforms complex segment exposure to values which can be compared with AC root-mean-square (rms) magnitude/duration curves, for determination of VF risk. Human contact could occur for any given time duration within the segment. The current of most risk is the greatest found for all possible instances of that duration. This is termed the "Probable Current" (PC) for that duration. All possible exposure durations in the waveform segment are considered, giving a set of PCs, thus allowing the plotting of a PC curve. The PC set is compared with a criterion for VF risk, termed the Justified Current (JC) curve. The theory is presented. Demonstrations and examples are given. Code is shown for generating the PC curve. VF risk can be found for an RCW using the rigorous algorithm presented. The VF for RCWs has not been considered previously. A rigorous statement of a method for VF risk assessment allows extension from regular waveforms to RCWs.