In this work we present a machine learning approach that is able to classify 30 cardiac abnormalities from an arbitrary number of electrocardiogram (ECG) leads. Features extracted by a deep convolutional neural network are combined with hand-crafted features (demographic, morphological, and heart rate variability metrics) and fed into a multilayer perceptron. We employ an Asymmetric Loss (ASL) function, which enables the model to focus on hard, but under-represented, samples. To mitigate the issue of ground-truth mislabeling and to provide robustness, we investigate the use of a self-learning label correction method that iteratively estimates correct labels during training. Leaderboard results show our team SMS+1 achieved challenge scores of 0.57 0.58 0.57.56 0.57 for twelve, six, four, three, and two-lead, respectively. Our model maintains the same diagnostic potential on both standard twelve-lead ECGs and reduced-lead ECGs.
Objective assessment of emotions is a challenging task.In this research, a new approach for detection and quantification of four emotions (pleasure, sadness, anger, and joy) is introduced based on the 2D regression models of emotions.ECG signal of 70 healthy female volunteers was recorded while the participants were stimulated by four different colors for five minutes.Induced arousal and valence levels were identified using the Self-Assessment Manikin test.Based on extracted features from Poincare Plot of heart rate series, 1-the label of emotion is determined based on 2-D emotion models using a rulebased approach, and 2-the strength of emotion is calculated in percentages using proposed mathematical formula.The results confirm the findings of previous studies that showed that each color is related to one emotion.The mean square error for estimation of arousal and valence levels were 0.0393 and 0.0536, respectively.In emotion classification, sensitivity of 96.78%, specificity of 99.4%, and the accuracy of 98.75% were achieved.The proposed method for measuring emotions can be of great help to psychiatrists to identify mental disorders.Using this method, one can measure and express the strength of emotions individually and has a comparative criterion for emotions in different individuals.
Tachyarrhythmia detection through RR interval analysis could improve performance of monitoring devices.In this paper a Poincaré plot-based image approach is presented.Three cardiac rhythms were analyzed in this study: normal sinus rhythm (NSR), atrial fibrillation (AF) and atrial bigeminy (AB).Using different MIT-BIH databases, 27955, 3363 and 76 images were generated for NSR, AF and AB respectively using a 2-minute window with 50 % overlap.The 80 % of the data available for each rhythm was used to create a reference rhythm image atlas.The remaining 20 % was classified into one of the three categories using mutual information.The process was iterated 10 times, in which images used to construct the atlas and used to create the test set were randomly selected.AF was correctly classified 94.12 %±0.45,AB 72.00 %±11.24 and NSR 80.70 %±0.54.The results of the present study suggest that Poincaré plot-based image analysis is a promising path for classifying different rhythms using only ventricular activity.
Computational fluid dynamics represents a valuable non-invasive approach to determine and assess physically meaningful parameters in a complex fluid dynamics system represented by AF. The aim of this study was the design, development and test of different LA motion fields in AF on a patient-specific 3D anatomical model to clarify the influence of contraction models on LA hemodynamics. Since LA motion field is not available from clinical data during AF, three displacement models were designed to simulate the irregular, disorganized, very rapid and strongly reduced LA contraction, a random model, a discrete random model and a con-tinuous sinusoidal model. Blood velocity fields, kinetic energy, vortex structures and blood stasis were analyzed in both SR and AF conditions in the LA. Velocities in SR were higher than in AF, particular-ly during atrial systole. The three AF models resulted in different wash-out velocities both at the mitral valve and at the ostium of the LA appendage. Vortices were also differently distributed inside the LA showing a more organized flow in the sinusoidal model which was also characterized by the lowest blood stasis (8.6%) inside the LA appendage. Overall, different LA de-formation models in AF affect LA hemodynamics and additional studies should be performed to develop a realistic contraction model to simulate AF episodes.
Early prediction of sepsis is of utmost importance to provide optimal care at an early stage. This work aims to use machine learning for early prediction of sepsis using ratio and power-based feature transformation. The feature transformation and feature selection process is optimized by applying a genetic algorithm (GA) based approach to extract the information specific to the sepsis from the given raw patient covariates that maximizes the underlying classification performance in terms of utility score. The proposed method begins with filling the missing values in the training dataset. Then, GA is applied strategically to identify influential ratio and power-based features from the raw patient covariates. The utility score is maximized as an objective of the optimization. RusBoost is used with default settings for underlying classification during optimization. Subsequently, an optimal RusBoost model is developed with a set of 55 identified features. Independent performance evaluation of the proposed method with the 2019 PhysioNet/CinC Challenge dataset has officially achieved 19th rank with a utility score of 30.9% on the full hidden test data. This work appears as Shivpatidar on the leader-board. The proposed early warning system has potential clinical value in critical care clinics.
Recently, a number of RR interval rhythm-based atrial fibrillation (AF) features have been developed and can achieve high classification accuracy for identifying AF from normal rhythm ECGs on clean signals. However, in dynamic ECG monitoring scenario, accurate location for QRS complexes is challenging, resulting in a deviation between calculated and reference RR interval sequences. This deviation can cause the failure of AF detection but its effect has not been quantified. This study addressed this concern and analyzed the anti-interference abilities of 14 commonly used AF features. Two types of Experiment were performed on the MIT-BIH AF database to simulate the deviation caused by QRS detection: 1) randomly moving forward or back several (0-15) labeled QRS locations to stimulate false detections, and 2) randomly missing several (0-9) labeled QRS locations to stimulate missing QRS detections. And the change in classification accuracy was regarded as a measure for evaluating the anti-interference ability of each feature. The results showed that features of AFEv, MAD, NFEn, COSEn and minRR showed high anti-interference abilities in both Experiments.
On a yearly basis, sepsis costs US hospitals more than any other health condition.A majority of patients who suffer from sepsis are not diagnosed at the time of admission.Early detection and antibiotic treatment of sepsis are vital to improve outcomes for these patients, as each hour of delayed treatment is associated with increased mortality.In this study our goal is to predict sepsis 12 hours before its diagnosis using vitals and blood tests routinely taken in the ICU.We have investigated the performance of several machine learning algorithms including XGBoost, CNN, CNN-LSTM and CNN-XGBoost.Contrary to our expectations, XGBoost outperforms all of the sequential models and yields the best hour-by-hour prediction, perhaps due to the way we imputed missing values, losing signal that relates to the time-series nature of the EHR data.We added feature engineering to detect change points in tests and vitals, resulting in 5% improvement in XGBoost.Our team, USF-Sepsis-Phys, achieved a utility score of 0.22 (untuned threshold) and an average of the three reported AUCs (test sets A, B, C) of 0.82.As expected with this AUC, the same model with tuned threshold (not run in the PhysioNet challenge) performed significantly better, as evaluated with 3-fold cross-validation of the entire PhyisoNet training set.
The aim of this preliminary study is to look how maternal-fetal heart rates and their beat to beat coupling patterns (λ are influenced by injection of β blocker(propranolol) into pregnant mice. Total of 9 fetuses from 9 pregnant female mice type of C57BL/6J were divided into three groups [control (3), β blockade (3) (4 mg), β blockade (3) (10 mg)]. On 17.5-day beat-to-beat maternal and fetal heart rates (MHR and FHR) were simultaneously measured for 20 minutes (10 minutes under normal condition and 10 minutes with saline (to control group) and propranolol (to the β blockade groups) solution by using an invasive maternal and fetal electrocardiogram techniques with needle electrodes. Results show that FHR decrease (p<0.05 for 10 mg) and maternal-fetal heart rate coupling (λ) patterns increases for 10 mg; p<0.05 and no significant changes for 4mg] followed by propranolol infusion (no change with saline). These changes could be caused by propranolol (dosage of 10 mg) which could have transmitted through the placental barrier. The presented results allow for assessment of the dosages of β adrenergic control of maternal and fetal heart, which will further enhance the value of the mouse as a model of heritable human pregnancy and hypertension.
The Purkinje network (PN) gains more clinically importance as it becomes target for pacing in rate control and defibrination. However, our understanding of the PN morphology arises from animal experiments, which might not transfer to humans. Therefore, we propose an automated computer simulation predicting physiological PN morphologies depending on the heart shape. It starts by generating virtual heart shapes from a statistical shape atlas and generates virtual PNs on the endocardial surface. For the combined virtual models the eikonal equation is solved to estimate the local activation times throughout the myocardium, which then feed forward to an simulation of the 12-lead surface ECG. From the simulated ECG the QRS-complex is compared against a healthy standard QRS-complex, which allows to estimate how physiological a PN morphology is.In our model, only bundle branch bifurcation points near the base or near the apex result in physiological QRS wave forms. For the right bundle, more physiological QRS waves can be obtained when the branching point is at the apex. Only a minor dependency of the ECG on the heart shape is found. However, a strong correlation between the bundle branch bifurcation points themselves is observed.
Left atrial appendage (LAA) closure is performed in atrial fibrillation (AF) patients to help prevent stroke. LAA closure using an occlusion implant is performed under imaging guidance. However, occlusion can be a complicated process due to the highly variable and heterogeneous LAA shapes across patients. Patient-specific implant selection and insertion processes are keys to the success of the procedure, yet subjective in nature. A population study of the angle of entry at the interatrial septum relative to the appendage can assist in both catheter design and patient-specific implant choice. In our population study, we analyzed the inherent clusters of the angles that were obtained between the septum normal and the LAA ostium plane. The number of inherent angle clusters matched the LAA four morphological classifications reported in the literature. Further, our exploratory analysis revealed that the normal from the ostium plane does not intersect the septum in all the samples under study. The insights gained from this study can help assist in making objective decisions during LAA closure.
Aortic pressure (P a ) waveforms are important for diagnosis of cardiovascular disease. However, the direct measurement of P a is invasive and expensive. In the paper, a new simplified Kalman filter (SKF) algorithm for blind system identification was employed for the reconstruction of P a waveforms using two peripheral artery pressure waveforms. The data of P a waveforms are collected from 24 human subjects. Simultaneously, brachial artery and femoral artery pressure waveforms data are generated from the simulation of a known two-channel finite impulse response system. In order to study the performance of the proposed SKF algorithm, different amounts of signal-to-noise ratio of the output signal were used in the experiment. Experimental results demonstrated that the proposed SKF algorithm had advantages in comparison with the canonical correlation analysis (CCA) algorithm. It is notable that the proposed SKF algorithm works much more noise-robust than the CCA algorithm in a wide range of SNR.
Identifying the site of origin (SOO) of outflow tract ventricular arrhythmias (OTVAs) is key to plan radiofrequency ablation procedures. Currently, electrophysiologist try to extract that information pre-operatively from the ECG, and intraoperatively from electroanatomical maps. In this work, we study the prediction of the SOO by using machine learning approaches trained with patient-specific electrophysiological simulations of subjects that suffer OTVA. We built patient-specific models for 11 patients with OTVA, including an enhance description of the myofiber orientation in the outflow tracts, and simulated the sequence of activation, the BSPM and ECG. Following, we triggered the arrhythmia from twelve different SOO. Simulations were in agreement with real ECGs, hence we used the simulated ECGs to train our machine learning algorithms and classify the different SOO. According to our results, V3 lead provides useful information for SOO localization. Obtained classification rates show that simulated ECGs can help to determine right ventricle versus left ventricle outflow tract origin.
Non-invasive fetal electrocardiography has the potential of providing vital information for evaluating the health status of the fetus.However, the low signal-tonoise ratio of the fetal electrocardiogram (ECG) impedes the applicability of the method in clinical practice.Residual noise in the fetal ECG, after the maternal ECG is suppressed, is often non-stationary, complex and has spectral overlap with the fetal ECG.We present a deep fully convolutional encoder-decoder framework, for removing the residual noise from single-channel fetal ECG.The method was tested in a broad simulated fetal ECG dataset with varying amount of noise.The results demonstrate that after the denoising there was an average increase in the correlation coefficient between the corrupted signals and the original ones from 0.6 to 0.8.Moreover, the suggested framework successfully handled different levels of noises in a single model.The network was further tested on real signals showing substantial noise removal performance, thus providing a promising approach for fetal ECG signal denoising.The presented method is able to significantly improve the quality of the extracted fetal ECG signals, having the advantage of preserving beat-to-beat morphological variations.
Chagas disease American trypanosomiasis is caused by a flagellated parasite: Trypanosoma cruzi, transmitted by an insect of the genus Triatoma and also by blood transfusions. In Latin America, the number of infected people is approximately 6 million, with a population exposed to the risk of infection of 550000. It is our interest to develop a non-invasive and low-cost methodology, capable of detecting any early cardiac alteration that also allows us to see dysautononia or dysfunction within 24 hours and with this it could be used to detect any cardiac alteration caused by T early Cruzi. For this, we analyzed the 24- hour Holter ECG records in 107 patients with ECG abnormalities (CH2), 102 patients without ECG alterations (CH1) who had positive serological results for Chagas disease and 83 volunteers without positive serological results for Chagas disease (CONTROL). Approximate entropy was used to quantify the regularity of electrocardiograms (ECG) in the three groups. We analyzed 288 ECG segments per patient. Significant differences were found between the CONTROL-CH1, CONTROL-CH2 and CH1- CH2 groups.
Bipolar ECG leads recorded from closely spaced electrodes are challenging in any context. When they are positioned distally with respect to the source field (far-field), the recovery of clinically useful signal content represents an even greater challenge. Due to the increased interest in ambulatory wellness devices, particularly wrist-worn devices, there is a renewed interest in recovering ECG signals from distally located bipolar leads.In this study 10 bipolar leads were simultaneously recorded at various locations along the left arm. At the same time, a conventional proximal reading on the chest using Lead I was also recorded and stored. This process was repeated for 11 healthy subjects. ECGs were recorded for a period of approximately 6 minutes for each subject and sampled at a frequency of 2048 Hz. Wavelet-based filtering using Daubechies 4 wavelet decomposition and soft threshold was applied to each lead. QRS detection performance was assessed against Lead I for each subject. This investigation found that a lead positioned transversally (using BIS gelled electrodes) on the upper arm provided the best accuracy against the benchmark QRS detection (SEN = 0.998, PPV = 0.984). The most distally positioned bipolar lead using dry electrodes faired least favourable (SEN = 0.272, PPV = 0.202).
The aim of this work is to investigate the relation between a phenomenon called "U-patterns" and their possible correlation to movement events in the context of sleep deprivation. U-patterns take place in the RR-interval time series during sleep. As their name suggests, these patterns present a U-shaped decrease-increase in RR-intervals, with a duration lasting from 20 to 40 seconds together with a minimum decrease of 15% in the local RR mean value.Over a span of 17 days, 15 healthy subjects (7males, 22.1 ± 1.7 yrs.) participated in a study of three subsequent stages. First, a baseline phase of seven days, during which the subjects slept normally. Immediately after, a sleep deprivation phase with a duration of three days, during which participants slept only three hours per night. Finally, in a 7-day recovery phase subjects went back to their normal baseline sleeping routine. Subjects underwent polysomnography (PSG) data acquisition while sleeping. U-patterns were extracted from RR-intervals while movement events were extracted from different PSG channels. Their relative temporal layout was studied to determine whether U-patterns are caused due to subject movement during sleep or vice versa. Results show that U-pattern/movement events are correlated, always initiated by U-patterns with movement events terminating before the termination of their respective U-patterns.
Heterozygous carriers of the A414G mutation in the HCN4 gene, which encodes the HCN4 protein, show moderate to severe sinus bradycardia.Tetramers of HCN4 subunits constitute the ion channels that conduct the cardiac hyperpolarization-activated 'funny current' (I f ), which plays an important modulating role in the pacemaker activity of sinus node cells.We assessed the mechanism by which the A414G mutation in HCN4 causes sinus bradycardia.We carried out voltage clamp experiments on HCN4 channels expressed in Chinese hamster ovary (CHO) cells and incorporated the experimentally observed mutationinduced changes in If into the Fabbri-Severi model of a single human sinus node cell.In the Fabbri-Severi model, the experimentally observed effects on I f increased the cycle length from 813 to 1004 ms, corresponding with a 19% decrease in beating rate from 74 to 60 beats/min.These mutation effects became more prominent at 10 nM ACh (vagal tone) and in the presence of a hyperpolarizing atrial load.We conclude that the experimentally identified mutation-induced changes in I f can explain the clinically observed sinus bradycardia in carriers of the A414G mutation in the HCN4 gene.
Transesophageal high-intensity focused ultrasound (HIFU) energy can be used to treat cardiac arrhythmia efficiently non-invasively.Since the esophagus is located just behind the heart, it offers a perfect acoustic window.Hence, HIFU can be directed toward the heart to perform ablation.In a previous study a HIFU probe with one 2D US image perpendicular to the esophagus axis for guidance purpose has been proposed.A new dual-mode HIFU probe with two perpendicular 2D US imaging plane is now under development.In this paper we propose a therapy guidance system, based on an intensity-based registration of the two perpendicular 2D US to preoperative 3D CT.As a proof of concept we developed the following evaluation framework on a numerical phantom: 1) we define a ground truth (GT) initial pose inside a CT volume and simulated two perpendicular US images from the CT data; 2) we run the registration framework from 55 randomly defined pose initializations around the initial GT pose; and 3) we estimated the accuracy of the registration by (a) the transformation parameter estimation errors and (b) Target Registration Error (TRE).The accuracy of the registration using two 2D US plane has been compared to the previous work performed on only one US plane.An improvement was observed when using two 2D US planes with regards to the previous one US plane.
Regulatory synchronization between the heart and the arterial walls is essential for optimal blood delivery to tissues. We investigated functional coherence between heart rhythm and aortic wall compliance in 30 volunteers aged 65 – 74. ECG, carotid and iliac pulse-wave were recorded and digitized at 2 kHz. Carotid-femoral pulse-wave transit time (cfTT) which reflects aortic compliance was assessed using the intersecting tangent algorithm at time-point of the maximal upstroke of the second derivative of the filtered pulse signal. Time-series of 256 heart cycles were used for analysis of heart rate variability (HRV) and cfTT variability analysis. Averaged power spectral density (aPSD) was estimated within selected frequency ranges by fast Fourier transform (FFT) approach. Magnitude squared coherence (MSC) between the both spectra was estimated. All volunteers exhibited variable temporal patterns of both HRV and cfTT variability. aPSD was reduced along with decreasing of time-window from 60 to 240 s. High coherent states between HRV and cfTT variability were observed as a short time prominent MSC peaks in almost all participants. Individual patterns of irregular MSC changes in time but not a generalized model of its fluctuations seem to reflect dynamic functional interaction between the heart and aortic compliance at an advanced age.
The enhancement of the late sodium current (I NaL ) has been demonstrated to contribute to the cardiac arrhythmias. However, its arrhythmogenic mechanism at the cellular and tissue level remains incompletely elucidated. In this study, the O’Hara-Rudy model of human ventricular cells was implemented for multi-level simulations. At the cellular level, the influences of the pathological enhanced I NaL on cardiac action potential characteristics, ion currents, intracellular concentration homeostasis and action potential duration (APD) restitution properties were simulated and analyzed. At the tissue level, a heterogeneous one dimensional (1D) strand was constructed to find out the impact of enhanced I NaL on APD dispersion and the vulnerable windows (VWs). The simulations revealed the role of augmenting I NaL in prolonging the APD, steepening the APD restitution curves, increasing the heterogeneity of the tissue and widening the VWs. Our simulation data provides a detailed mechanistic insight into the pro-arrhythmic role of the enhanced I NaL at both cellular and tissue levels.
