Multiresolution decomposition of the signal-averaged ECG using the mallat approach for prediction of arrhythmic events after myocardial infarction

Longitudinally tracking personal physiomes for precision management of childhood epilepsy

Our current understanding of human physiology and activities is largely derived from sparse and discrete individual clinical measurements. To achieve precise, proactive, and effective health management of an individual, longitudinal, and dense tracking of personal physiomes and activities is required, which is only feasible by utilizing wearable biosensors. As a pilot study, we implemented a cloud computing infrastructure to integrate wearable sensors, mobile computing, digital signal processing, and machine learning to improve early detection of seizure onsets in children. We recruited 99 children diagnosed with epilepsy and longitudinally tracked them at single-second resolution using a wearable wristband, and prospectively acquired more than one billion data points. This unique dataset offered us an opportunity to quantify physiological dynamics (e.g., heart rate, stress response) across age groups and to identify physiological irregularities upon epilepsy onset. The high-dimensional personal physiome and activity profiles displayed a clustering pattern anchored by patient age groups. These signatory patterns included strong age and sex-specific effects on varying circadian rhythms and stress responses across major childhood developmental stages. For each patient, we further compared the physiological and activity profiles associated with seizure onsets with the personal baseline and developed a machine learning framework to accurately capture these onset moments. The performance of this framework was further replicated in another independent patient cohort. We next referenced our predictions with the electroencephalogram (EEG) signals on selected patients and demonstrated that our approach could detect subtle seizures not recognized by humans and could detect seizures prior to clinical onset. Our work demonstrated the feasibility of a real-time mobile infrastructure in a clinical setting, which has the potential to be valuable in caring for epileptic patients. Extension of such a system has the potential to be leveraged as a health management device or longitudinal phenotyping tool in clinical cohort studies.

Monitoring respiratory rate based on tracheal sounds. First experiences

The objective was to develop a non-invasive method for continuously monitoring respiratory rate (RR) based on tracheal sounds. 25 volunteers and 36 patients with chronic pulmonary diseases were enrolled in a clinical study. Tracheal sounds were acquired using a contact piezoelectric sensor placed on the examinee's throat and analyzed using a combined investigation of the sound envelope and frequency content. RR estimates were compared to reference measurements taken from a pneumotachometer coupled to a face mask worn by the examinee. RR was also manually counted by a respiratory technician. Two types of breathing (mouth and nose) and three different positions were studied (fowler, semi-fowler and supine). RR estimated in volunteers had a success rate (SR) of 96%, a correlation coefficient (r) of 0.99 and a standard error of the estimate (SEE) of 0.56. The RR estimated in patients was comparable or slightly better (SR = 85%, r = 0.93 and SEE = 1.49) than those obtained by manual count (SR = 82%, r = 0.91, SEE = 1.58), which is the method widely used in clinical settings. No significant difference in the capacity to estimate RR was found related to posture and breathing type, making this method useful for continuous monitoring.

Wigner-Ville analysis and classification of electrocardiograms during thrombolysis

Non-stationary analysis of electrocardiograms (ECGs) using Wigner-Ville distribution is presented. Analysis was performed on subjects with acute myocardial infarction who had undergone thrombolysis, in Holter recordings of lead V1. The distinction between successfully and non-successfully thrombolysed patients was evaluated, based on time-frequency features of the Wigner-Ville transformed ECGs at the sixth hour after lysis. Characteristic parameters were extracted from time-frequency areas, and linear discriminant analysis was performed on these parameters, leading to a prediction index to distinguish the two classes. Thirteen features were found statistically significant by t-test and were used for the classification with linear modelling. Out of these features, four corresponded to frequencies lower than 25 Hz and higher than 50 Hz for, roughly, the QRS complex, five features corresponded to all the frequency bands of, roughly, the ST area, and the last four features corresponded to the T-wave. The feature-vector used in linear modelling was iteratively generated, and the iterative prediction found all 18 features significant. The iterative method resulted in better classification than that of the standard statistical procedure (3.8% error against 18.1% with the classic method). The evolution of the prediction index with time for the first 12 h was different for the successfully and non-successfully thrombolysed groups. Specifically, in the successful thrombolysis group, oscillations and variation with time were more obvious, indicating a possible difference in the dynamics of the cardiac system.

Increased cycle length variability during ventricular fibrillation: a novel predictor of arrhythmia recurrence

To evaluate the clinical value of cycle length (CL) variability during ventricular fibrillation (VF), 26 patients who underwent implantable cardioverter defibrillator (ICD) implantation were enrolled. In VF induced for defibrillation testing, mean and SD of VFCL, mean successive differences (MSD) of VFCL, and coefficient of variations of the VFCL (CV(FF)) (SD x 100/mean VFCL) were calculated. During the follow-up period of 20 +/- 2 months, ventricular arrhythmias recurred in 13 patients. MSD and CV(FF) were 31 +/- 3(*) ms and 15.6 +/- 1.3(**) in recurrence group (n = 13), and 17 +/- 2 ms and 9.0 +/-1.1 in non-recurrence group (n = 13) ((*)P <.005, (**)P <.001 vs. nonrecurrence group). Relatively good repeatability of mean VFCL, MSD and CV(FF) in each patient was confirmed by the Bland-Altman method. In VF induced by programmed ventricular stimulation before ICD implantation, MSD and CV(FF) in recurrence group were also increased significantly. Kaplan-Meier estimates revealed that MSD >or= 20 ms and CV(FF) >or= 12 predicted higher arrhythmia recurrence (MSD, P =.039; CV(FF), P =.0069 by the log-rank test). By multivariate analysis, CV(FF) >or= 12 was a significant predictor of recurrent arrhythmic events (P =.019). In conclusion, the CL variability of VF, which was evaluated as MSD and CV(FF), is increased in patients with arrhythmia recurrence. These values may reflect the degree of electrical heterogeneity, and appears to be useful indexes of the future arrhythmic events.

Utility of current risk stratification tests for predicting major arrhythmic events after myocardial infarction

OBJECTIVES

We surveyed the literature to estimate prediction values for five common tests for risk of major arrhythmic events (MAEs) after myocardial infarction. We then determined feasibility of a staged risk stratification using combinations of noninvasive tests, reserving an electrophysiologic study (EPS) as the final test.

BACKGROUND

Improved approaches are needed for identifying those patients at highest risk for subsequent MAE and candidates for implantable cardioverter-defibrillators.

METHODS

We located 44 reports for which values of MAE incidence and predictive accuracy could be inferred: signal-averaged electrocardiography; heart rate variability; severe ventricular arrhythmia on ambulatory electrocardiography; left ventricular ejection fraction; and EPS. A meta-analysis of reports used receiver-operating characteristic curves to estimate mean values for sensitivity and specificity for each test and 95% confidence limits. We then simulated a clinical situation in which risk was estimated by combining tests in three stages.

RESULTS

Test sensitivities ranged from 42.8% to 62.4%; specificities from 77.4% to 85.8%. A three-stage stratification yielded a low-risk group (80.0% with a two-year MAE risk of 2.9%), a high-risk group (11.8% with a 41.4% risk) and an unstratified group (8.2% with an 8.9% risk equivalent to a two-year incidence of 7.9%).

CONCLUSIONS

Sensitivities and specificities for the five tests were relatively similar. No one test was satisfactory alone for predicting risk. Combinations of tests in stages allowed us to stratify 91.8% of patients as either high-risk or low-risk. These data suggest that a large prospective study to develop a robust prediction model is feasible and desirable.

Signal averaging in Wolff-Parkinson-White syndrome: evidence that fractionated activation is not necessary for body surface high-frequency potentials

It is commonly assumed that the presence of high frequency components in body surface potentials implies that fractionated activation fronts, caused by heterogeneously viable tissue, are present in the heart. However, it is possible that non-fractionated activation fronts can also give rise to high frequency surface potentials and that the relative amount of high frequency power is related to the complexity of the activation sequence. In a test of this idea, averaged body surface potentials were recorded during the entire QRS complex of nine Wolff-Parkinson-White (WPW) patients in situations in which fractionated activation fronts should not have been present, but which represent increasing degrees of complexity of ventricular activation: (1) postoperative ectopic pacing from subepicardial wires placed during surgery, when a single coherent activation front was present throughout most of the QRS; (2) Preoperative preexcited rhythm, when a single coherent activation front was present for one portion of the QRS (the delta wave); and (3) postoperative normal rhythm, when two or more activation fronts were present in the ventricles throughout most of the QRS. For comparison, averaged body surface potentials were also analyzed during the last 40 ms of the QRS complex and the ST segment of 14 postinfarction patients with chronic ventricular tachycardia. In the patients with WPW syndrome, relatively high frequency content increased (attenuation -36.7 vs -27.2 vs -18.3 dB) and QRS width decreased (160.7 vs 125.9 vs 94.1 ms) significantly from paced to preoperative to postoperative beats. Significant high frequency content was present in all cases, showing that coherent activation fronts can give rise to high frequencies. Interestingly, the postoperative QRS of WPW patients contained a larger proportion of high frequency power than did the late potentials of the patients with ventricular tachycardia. Thus, while the presence of late fractionated body surface potentials may be a marker for ventricular tachycardia, these potentials by themselves do not necessarily signify that the underlying cardiac activation giving rise to these signals is fractionated.