A Crowdsourced AI Framework for Atrial Fibrillation Detection in Apple Watch and Kardia Mobile ECGs.

Background: Atrial fibrillation (AFib) detection via mobile ECG devices is promising, but algorithms often struggle to generalize across diverse datasets and platforms, limiting their real-world applicability. Objective: This study aims to develop a robust, generalizable AFib detection approach for mobile ECG devices using crowdsourced algorithms. Methods: We developed a voting algorithm using random forest, integrating six open-source AFib detection algorithms from the PhysioNet Challenge. The algorithm was trained on an AliveCor dataset and tested on two disjoint AliveCor datasets and one Apple Watch dataset. Results: The voting algorithm outperformed the base algorithms across all metrics: the average of sensitivity (0.884), specificity (0.988), PPV (0.917), NPV (0.985), and F1-score (0.943) on all datasets. It also demonstrated the least variability among datasets, signifying its highest robustness and effectiveness in diverse data environments. Moreover, it surpassed Apple's algorithm on all metrics and showed higher specificity but lower sensitivity than AliveCor's Kardia algorithm. Conclusions: This study demonstrates the potential of crowdsourced, multi-algorithmic strategies in enhancing AFib detection. Our approach shows robust cross-platform performance, addressing key generalization challenges in AI-enabled cardiac monitoring and underlining the potential for collaborative algorithms in wearable monitoring devices.

Electrocardiographic measures of repolarization heterogeneity are not predictive for Torsades de Pointes among undifferentiated patients with prolonged QTc: A case control study.

INTRODUCTION: Torsades de Pointes (TdP) is a potentially lethal polymorphic ventricular tachydysrhythmia associated with and caused by prolonged myocardial repolarization. However, prediction of TdP is challenging. We sought to determine if electrocardiographic myocardial repolarization heterogeneity is necessary and predictive of TdP. METHODS: We performed a case control study of TdP at a large urban hospital. We identified cases based on a hospital center electrocardiogram (ECG) database search for tracings from 1/2005 to 6/2019 with heart rate corrected QT (QTc) > 500, QRS < 120, and heart rate (HR) < 60, and a subsequent natural language search of electronic health records for the terms: TdP, polymorphic ventricular tachycardia, sudden cardiac death, and relevant variants. Controls were drawn in a 2:1 ratio to cases from a similar pool of ECGs, and matching for QTc, heart rate, sex, and age. We abstracted historical, laboratory, and ECG data using detailed written instructions and an electronic database. We included a second blinded data abstractor to test data abstraction and manual ECG measurement reliability. We used General Electric (GE) QT Guard software for automated repolarization measurements. We compared groups using unpaired statistics. RESULTS: We included 75 cases and 150 controls. The number of current QTc prolonging medications and serum electrolytes were substantially the same between the two groups. We found no significant difference in measures of QT or T wave repolarization heterogeneity. CONCLUSION: Electrocardiographic repolarization heterogeneity is not greater in otherwise unselected patients with QTc prolongation who suffer TdP and does not appear predictive of TdP. However, previous observations suggest specific repolarization characteristics may be useful for defined patient subgroups at risk for TdP.

Machine learning models of 6-lead ECGs for the interpretation of left ventricular hypertrophy (LVH).

BACKGROUND: Left Ventricular Hypertrophy (LVH) is closely linked to the cardiovascular disease prognosis, and thus, timely diagnosis improves outcomes. Diagnosis is challenging due to dependency on doctor's visits and a 12­lead ECG. In addition, the interpretation of LVH from ECGs is challenging due to variability of ECG measurements, body habitus, electrode positioning, several LVH ECG criteria and EP mechanisms. The aims of this study are to evaluate different big data-driven machine learning models for ECG LVH interpretation based on limb leads only, and to compare the performance of an ECG parameter-based statistical model with a deep learning-based model. METHODS AND DATA: The first two models are binary class Random Forest (RF) models, an ensemble learning method which constructs many decision trees at training time and predicts the class chosen by the greatest number of trees at inference time. One random forest is trained using the following five features: lead aVL R-wave amplitude, lead I, II, aVL ST segment amplitude, and QRS duration. The second RF model uses 54 features across all limb leads, including the five features used by the smaller model. The second type of model is a multi-class deep neural network (DNN) which takes median beats of 6 limb leads arranged in Cabrera sequence as input. The signal preprocessing included forming median beats, filtering with a 40-Hz lowpass filter, and down-sampling to 125 Hz. The DNN models consist of 1 lead-formation convolutional layer, 5 downsampling convolutional resnet blocks with skip connections, and 3 fully connected layers. The training dataset consisted of 1 million 10-s 12­lead ECGs, and an independent test dataset consisted of 250,000 10-s ECGs from the Mayo Clinic. RESULTS: The five-parameter RF model has the prediction performance of Area Under the Receiver-Operator Curve (AUC) 0.78, and the larger RF model had AUC of 0.83. The DNN model for ECG LVH detection achieves AUC 0.92 using only the limb leads, compared to an AUC of 0.98 for the full 12­lead DNN. CONCLUSION: The study shows that machine learning models trained only on limb leads achieve promising results with potential to add clinical value to early detection mechanisms. We also observe that the RF model splits parameters by thresholds known to be characteristic of LVH, and that the DNN model can automatically detect morphology differences from 6 limb lead ECGs. This will be meaningful for expanding the capabilities of potential electrical LVH detection in mobile 6­lead ECG devices.

Age, sex and race bias in automated arrhythmia detectors.

Despite the recent explosion of machine learning applied to medical data, very few studies have examined algorithmic bias in any meaningful manner, comparing across algorithms, databases, and assessment metrics. In this study, we compared the biases in sex, age, and race of 56 algorithms on over 130,000 electrocardiograms (ECGs) using several metrics and propose a machine learning model design to reduce bias. Participants of the 2021 PhysioNet Challenge designed and implemented working, open-source algorithms to identify clinical diagnosis from 2- lead ECG recordings. We grouped the data from the training, validation, and test datasets by sex (male vs female), age (binned by decade), and race (Asian, Black, White, and Other) whenever possible. We computed recording-wise accuracy, area under the receiver operating characteristic curve (AUROC), area under the precision recall curve (AUPRC), F-measure, and the Challenge Score for each of the 56 algorithms. The Mann-Whitney U and the Kruskal-Wallis tests assessed the performance differences of algorithms across these demographic groups. Group trends revealed similar values for the AUROC, AUPRC, and F-measure for both male and female groups across the training, validation, and test sets. However, recording-wise accuracies were 20% higher (p < 0.01) and the Challenge Score 12% lower (p = 0.02) for female subjects on the test set. AUPRC, F-measure, and the Challenge Score increased with age, while recording-wise accuracy and AUROC decreased with age. The results were similar for the training and test sets, but only recording-wise accuracy (12% decrease per decade, p < 0.01), Challenge Score (1% increase per decade, p < 0.01), and AUROC (1% decrease per decade, p < 0.01) were statistically different on the test set. We observed similar AUROC, AUPRC, Challenge Score, and F-measure values across the different race categories. But, recording-wise accuracies were significantly lower for Black subjects and higher for Asian subjects on the training (31% difference, p < 0.01) and test (39% difference, p < 0.01) sets. A top performing model was then retrained using an additional constraint which simultaneously minimized differences in performance across sex, race and age. This resulted in a modest reduction in performance, with a significant reduction in bias. This work provides a demonstration that biases manifest as a function of model architecture, population, cost function and optimization metric, all of which should be closely examined in any model.

Comparison of automated interval measurements by widely used algorithms in digital electrocardiographs.

BACKGROUND: Automated measurements of electrocardiographic (ECG) intervals by current-generation digital electrocardiographs are critical to computer-based ECG diagnostic statements, to serial comparison of ECGs, and to epidemiological studies of ECG findings in populations. A previous study demonstrated generally small but often significant systematic differences among 4 algorithms widely used for automated ECG in the United States and that measurement differences could be related to the degree of abnormality of the underlying tracing. Since that publication, some algorithms have been adjusted, whereas other large manufacturers of automated ECGs have asked to participate in an extension of this comparison. METHODS: Seven widely used automated algorithms for computer-based interpretation participated in this blinded study of 800 digitized ECGs provided by the Cardiac Safety Research Consortium. All tracings were different from the study of 4 algorithms reported in 2014, and the selected population was heavily weighted toward groups with known effects on the QT interval: included were 200 normal subjects, 200 normal subjects receiving moxifloxacin as part of an active control arm of thorough QT studies, 200 subjects with genetically proved long QT syndrome type 1 (LQT1), and 200 subjects with genetically proved long QT syndrome Type 2 (LQT2). RESULTS: For the entire population of 800 subjects, pairwise differences between algorithms for each mean interval value were clinically small, even where statistically significant, ranging from 0.2 to 3.6milliseconds for the PR interval, 0.1 to 8.1milliseconds for QRS duration, and 0.1 to 9.3milliseconds for QT interval. The mean value of all paired differences among algorithms was higher in the long QT groups than in normals for both QRS duration and QT intervals. Differences in mean QRS duration ranged from 0.2 to 13.3milliseconds in the LQT1 subjects and from 0.2 to 11.0milliseconds in the LQT2 subjects. Differences in measured QT duration (not corrected for heart rate) ranged from 0.2 to 10.5milliseconds in the LQT1 subjects and from 0.9 to 12.8milliseconds in the LQT2 subjects. CONCLUSIONS: Among current-generation computer-based electrocardiographs, clinically small but statistically significant differences exist between ECG interval measurements by individual algorithms. Measurement differences between algorithms for QRS duration and for QT interval are larger in long QT interval subjects than in normal subjects. Comparisons of population study norms should be aware of small systematic differences in interval measurements due to different algorithm methodologies, within-individual interval measurement comparisons should use comparable methods, and further attempts to harmonize interval measurement methodologies are warranted.

Optimal method of measuring the T-peak to T-end interval for risk stratification in primary prevention.

Aims: Several published investigations demonstrated that a longer T-peak to T-end interval (Tpe) implies increased risk for ventricular tachyarrhythmia (VT/VF) and mortality. Tpe has been measured using diverse methods. We aimed to determine the optimal Tpe measurement method for screening purposes. Methods and results: We evaluated 305 patients with LVEF ≤ 35% and an implantable cardioverter-defibrillator implanted for primary prevention. Tpe was measured using seven different methods described in the literature, including six manual methods and the automated algorithm '12SL', and was corrected for heart rate. Endpoints were VT/VF and death. To account for differences in the magnitude of Tpe measurements, results are expressed in standard deviation (SD) increments. We evaluated the clinical utility of each measurement method based on predictive ability, fraction of immeasurable tracings, and intra- and interobserver correlation. >Over 31 ± 23 months, 82 (27%) patients had VT/VF, and over 49 ± 21 months, 91 (30%) died. Several rate-corrected Tpe measurement methods predicted VT/VF (HR per SD 1.20-1.34; all P < 0.05), and nearly all methods (both corrected and uncorrected) predicted death (HR per SD 1.19-1.35; all P < 0.05). Optimal predictive ability, readability, and correlation were found in the automated 12SL method and the manual tangent method in lead V2. Conclusion: For the prediction of VT/VF, the utility of Tpe depends upon the measurement method, but for the prediction of mortality, most published Tpe measurement methods are similarly predictive. Heart rate correction improves predictive ability. The automated 12SL method performs as well as any manual measurement, and among manual methods, lead V2 is most useful.

Is myocardial repolarization duration associated with repolarization heterogeneity?

BACKGROUND: Dispersion of repolarization is theorized as one mechanism by which myocardial repolarization prolongation causes lethal torsades de pointes, (TdP). Our primary purpose was to determine whether prolongation of myocardial repolarization as measured by the heart rate-corrected J-to-T peak interval (JTpkc), is associated with repolarization heterogeneity as measured by transmural dispersion, defined as the median duration from the peak to the end of the T wave (TpTe). METHODS: A retrospective cohort study was performed at a single urban tertiary ED from July 2011-September 2012. Inclusion criteria included all consecutive ED patients with ECG based on QTc and QRS intervals. Automated measurements of all intervals were performed. The association of JTpkc with the dependent variable TpTe was assessed after adjustment for QRS and RR interval durations with a multiple linear regression model. A secondary analysis included a similar adjusted assessment of the association of JTpkc with QT dispersion, QTd. Finally, we constructed two multiple regression models to assess the association of clinical causative factors of TdP with TpTe and JTpkc. RESULTS: Eight hundred seventy-four cases were included: 186 with QTc <500 ms, 118 with QTc ≥500 and QRS ≥120 ms, and 570 with QTc ≥500 and QRS <120 ms. The coefficient for association of JTpkc with TpTe was -0.10 (95%CI -0.15 to -0.05), and for JTpkc with QTd was 0.03 (95% CI -0.01 to 0.06). Clinical causative TdP factors were associated more with JTpkc than TpTe. CONCLUSION: Repolarization duration as measured by JTpkc is not positively associated with dispersion of repolarization as measured by TpTe or QTd. Dispersion of repolarization may not be a critical mechanistic link between QTc prolongation and TdP.

Comparison of automated measurements of electrocardiographic intervals and durations by computer-based algorithms of digital electrocardiographs.

BACKGROUND AND PURPOSE: Automated measurements of electrocardiographic (ECG) intervals are widely used by clinicians for individual patient diagnosis and by investigators in population studies. We examined whether clinically significant systematic differences exist in ECG intervals measured by current generation digital electrocardiographs from different manufacturers and whether differences, if present, are dependent on the degree of abnormality of the selected ECGs. METHODS: Measurements of RR interval, PR interval, QRS duration, and QT interval were made blindly by 4 major manufacturers of digital electrocardiographs used in the United States from 600 XML files of ECG tracings stored in the US FDA ECG warehouse and released for the purpose of this study by the Cardiac Safety Research Consortium. Included were 3 groups based on expected QT interval and degree of repolarization abnormality, comprising 200 ECGs each from (1) placebo or baseline study period in normal subjects during thorough QT studies, (2) peak moxifloxacin effect in otherwise normal subjects during thorough QT studies, and (3) patients with genotyped variants of congenital long QT syndrome (LQTS). RESULTS: Differences of means between manufacturers were generally small in the normal and moxifloxacin subjects, but in the LQTS patients, differences of means ranged from 2.0 to 14.0 ms for QRS duration and from 0.8 to 18.1 ms for the QT interval. Mean absolute differences between algorithms were similar for QRS duration and QT intervals in the normal and in the moxifloxacin subjects (mean ≤6 ms) but were significantly larger in patients with LQTS. CONCLUSIONS: Small but statistically significant group differences in mean interval and duration measurements and means of individual absolute differences exist among automated algorithms of widely used, current generation digital electrocardiographs. Measurement differences, including QRS duration and the QT interval, are greatest for the most abnormal ECGs.

P-pulmonale and the development of atrial fibrillation.

BACKGROUND: P wave ≥0.25mV in inferior leads (P pulmonale) occurs in chronic lung diseases that underlie atrial fibrillation (AF). The purpose of this study was to elucidate the prognostic value of P pulmonale for development of AF. METHODS AND RESULTS: Digital analysis of 12-lead electrocardiogram (ECG) was conducted to enroll patients with P pulmonale from among a database containing 308,391 ECGs. In a total of 591 patients (382 men; 56.4±14.8 years) with P pulmonale (follow-up, 46.7±65.6 months), AF occurred in 61 patients (AF group), but did not occur in 530 patients (non-AF group). Male gender was significantly more prevalent in the AF group than in the non-AF group (80.3% vs. 62.8%, P=0.0047). P-wave duration and PQ interval were significantly longer in the AF group than in the non-AF group (115.4±17.2ms vs. 107.0±17.2ms, P=0.0003 and 166.3±23.9ms vs. 153.2±25.4ms, P=0.0001, respectively). In the total patient group, multivariate Cox proportional-hazards analysis confirmed that male gender (hazard ratio [HR], 2.24; 95% confidence interval [CI]: 1.02-5.49; P=0.045), PQ interval >150ms (HR, 6.89; 95% CI: 2.39-29.15; P<0.0001), and P-wave axis <74° (HR, 2.55; 95% CI: 1.20-5.41; P=0.016) were associated with AF development. In medication-free patients (n=400), only PQ interval >150ms (HR, 9.26; 95% CI: 1.75-170.65; P=0.0055) was independently and significantly associated with AF development. CONCLUSIONS: PQ interval is the strongest stratifier for AF development in P pulmonale.

How can computerized interpretation algorithms adapt to gender/age differences in ECG measurements?

It is well known that there are gender differences in 12 lead ECG measurements, some of which can be statistically significant. It is also an accepted practice that we should consider those differences when we interpret ECGs, by either a human overreader or a computerized algorithm. There are some major gender differences in 12 lead ECG measurements based on automatic algorithms, including global measurements such as heart rate, QRS duration, QT interval, and lead-by-lead measurements like QRS amplitude, ST level, etc. The interpretation criteria used in the automatic algorithms can be adapted to the gender differences in the measurements. The analysis of a group of 1339 patients with acute inferior MI showed that for patients under age 60, women had lower ST elevations at the J point in lead II than men (57±91µV vs. 86±117µV, p<0.02). This trend was reversed for patients over age 60 (lead aVF: 102±126µV vs. 84±117µV, p<0.04; lead III: 130±146µV vs. 103±131µV, p<0.007). Therefore, the ST elevation thresholds were set based on available gender and age information, which resulted in 25% relative sensitivity improvement for women under age 60, while maintaining a high specificity of 98%. Similar analyses were done for prolonged QT interval and LVH cases. The paper uses several design examples to demonstrate (1) how to design a gender-specific algorithm, and (2) how to design a robust ECG interpretation algorithm which relies less on absolute threshold-based criteria and is instead more reliant on overall morphology features, which are especially important when gender information is unavailable for automatic analysis.

Prognostic implications of progressive cardiac conduction disease.

BACKGROUND: Progressive cardiac conduction disease (PCCD), characterized by temporal increase in PR interval and QRS duration, may be attributed to diverse pathophysiological mechanisms. This study aimed to investigate whether PCCD is associated with increased risk of cardiovascular morbidity and mortality. METHODS AND RESULTS: Digital analysis of 12-lead ECG was performed to select patients with PCCD from among a database containing 359,737 ECGs. Long-term prognosis of PCCD was assessed in a large hospital-based population: 458 patients (341 males; mean age, 57.9 ± 14.7 years) with PCCD were enrolled. During a mean follow-up of 13.3 ± 6.4 years, 109 patients were hospitalized for heart failure (HF), and there were 16 and 59 deaths from cardiovascular diseases and all causes, respectively. Multivariate Cox proportional hazards analysis confirmed (1) a significant association of temporal incremental rate of PR interval (≥ 2 ms/year) and QRS duration (≥ 3 ms/year) with HF hospitalization (hazard ratio [HR], 2.34; 95% confidence interval [CI], 1.36-4.05; P=0.002 and HR, 2.08; 95% CI, 1.25-3.53; P=0.01, respectively) and (2) a significant association of temporal incremental rate of PR interval (≥ 4 ms/year) and QRS duration (≥ 5 ms/year) with cardiovascular mortality (HR, 6.9; 95% CI, 1.47-36.96; P=0.02 and HR, 4.31; 95% CI, 1.19-16.5; P=0.03, respectively). CONCLUSIONS: The severity of PCCD was independently and significantly associated with HF hospitalization and cardiovascular mortality.

The detection of T-wave variation linked to arrhythmic risk: an industry perspective.

Although the scientific literature contains ample descriptions of peculiar patterns of repolarization linked to arrhythmic risk, the objective quantification and classification of these patterns continues to be a challenge that impacts their widespread adoption in clinical practice. To advance the science, computerized algorithms spawned in the academic environment have been essential in order to find, extract and measure these patterns. However, outside the strict control of a core lab, these algorithms are exposed to poor quality signals and need to be effective in the presence of different forms of noise that can either obscure or mimic the T-wave variation (TWV) of interest. To provide a practical solution that can be verified and validated for the market, important tradeoffs need to be made that are based on an intimate understanding of the end-user as well as the key characteristics of either the signal or the noise that can be used by the signal processing engineer to best differentiate them. To illustrate this, two contemporary medical devices used for quantifying T-wave variation are presented, including the modified moving average (MMA) for the detection of T-wave Alternans (TWA) and the quantification of T-wave shape as inputs to the Morphology Combination Score (MCS) for the trending of drug-induced repolarization abnormalities.

Relationships between the T-peak to T-end interval, ventricular tachyarrhythmia, and death in left ventricular systolic dysfunction.

AIMS: The interval between the T-wave's peak and end (Tpe), an electrocardiographic (ECG) index of ventricular repolarization, has been proposed as an indicator of arrhythmic risk. We aimed to clarify the clinical usefulness of Tpe for risk stratification. METHODS AND RESULTS: We evaluated 327 patients with left ventricular ejection fraction (LVEF) ≤ 35% (75% male, LVEF 23 ± 7%). All patients had an implanted implantable cardioverter-defibrillator (ICD). Clinical data and ECGs were analysed at baseline. Prospective follow-up for the endpoints of appropriate ICD therapy and mortality was conducted via periodic device interrogation, chart review, and the Social Security Death Index. During device clinic follow-up of 17 ± 12 months, 59 (18%) patients had appropriate ICD therapy, and during mortality follow-up of 30 ± 13 months, 67 (21%) patients died. A longer Tpe(c) predicted appropriate ICD therapy, death, and the combination of appropriate ICD therapy or death (P< 0.01 for each endpoint). On multivariable analysis correcting for other univariable predictors, Tpe(c) remained predictive of ICD therapy [hazard ratio (HR) per 10 ms increase: 1.16, P= 0.02], all-cause mortality (HR per 10 ms: 1.14, P= 0.03), and the composite endpoint of ICD therapy or death (HR per 10 ms: 1.16, P< 0.01). CONCLUSIONS: In patients with left ventricular systolic dysfunction and an implanted ICD, Tpe(c) independently predicts both ventricular tachyarrhythmia and overall mortality.

Risk determinants in individuals with a spontaneous type 1 Brugada ECG.

BACKGROUND: Spontaneous coved ST-segment elevation ≥2 mm followed by a negative T-wave in the right precordial leads (type 1 Brugada ECG) is diagnostic of Brugada syndrome (BS), but there is a false-positive rate. METHODS AND RESULTS: Computer-processed analysis of a 12-lead ECG database containing 49,286 females and 52,779 males was performed to select patients with a spontaneous type 1 Brugada ECG for an examination of the association of this ECG characteristic with long-term prognosis. There were 185 patients with a spontaneous type 1 Brugada ECG and of these, 16 (15 males; mean age, 46.7±14.0 years) were diagnosed with BS and 15 patients (all males; mean age, 50.1±13.4 years) were undiagnosed. The PQ interval was significantly longer in the diagnosed patients than in the undiagnosed patients (187.4±28.3 ms vs. 161.2±21.5 ms; P=0.0073). The T-wave in lead V(1) was more negative in the diagnosed patients than in the undiagnosed patients (-170.2±174.6 µV vs. -43.2±122.3 µV, P=0.027). Multivariate analysis revealed that a PQ interval ≥170 ms and T-wave amplitude <105 µV in lead V(1) were independent risk stratifiers of life-threatening events. Survival analysis (mean follow-up, 78.6±81.8 months) showed that the PQ interval and a negative T-wave in lead V(1) were significantly associated with poor prognosis. CONCLUSIONS: Analysis of a standard 12-lead ECG can stratify the prognosis of patients with a spontaneous type 1 Brugada ECG.

Atrial fibrillation detection in outpatient electrocardiogram monitoring: An algorithmic crowdsourcing approach.

BACKGROUND: Atrial fibrillation (AFib) is the most common cardiac arrhythmia associated with stroke, blood clots, heart failure, coronary artery disease, and/or death. Multiple methods have been proposed for AFib detection, with varying performances, but no single approach appears to be optimal. We hypothesized that each state-of-the-art algorithm is appropriate for different subsets of patients and provides some independent information. Therefore, a set of suitably chosen algorithms, combined in a weighted voting framework, will provide a superior performance to any single algorithm. METHODS: We investigate and modify 38 state-of-the-art AFib classification algorithms for a single-lead ambulatory electrocardiogram (ECG) monitoring device. All algorithms are ranked using a random forest classifier and an expert-labeled training dataset of 2,532 recordings. The seven top-ranked algorithms are combined by using an optimized weighting approach. RESULTS: The proposed fusion algorithm, when validated on a separate test dataset consisting of 4,644 recordings, resulted in an area under the receiver operating characteristic (ROC) curve of 0.99. The sensitivity, specificity, positive-predictive-value (PPV), negative-predictive-value (NPV), and F1-score of the proposed algorithm were 0.93, 0.97, 0.87, 0.99, and 0.90, respectively, which were all superior to any single algorithm or any previously published. CONCLUSION: This study demonstrates how a set of well-chosen independent algorithms and a voting mechanism to fuse the outputs of the algorithms, outperforms any single state-of-the-art algorithm for AFib detection. The proposed framework is a case study for the general notion of crowdsourcing between open-source algorithms in healthcare applications. The extension of this framework to similar applications may significantly save time, effort, and resources, by combining readily existing algorithms. It is also a step toward the democratization of artificial intelligence and its application in healthcare.

A Novel High-Resolution Surface Electrocardiographic Method to Identify and Characterize Myocardial Scar: A Proof-of-Concept Study.

BACKGROUND: The placement of the left ventricular (LV) lead in an area free of myocardial scar is an important determinant of cardiac resynchronization therapy response. We sought to develop and validate a simple, practical, and novel electrocardiographic (ECG)-based approach to intraoperatively identify the presence of LV scar. We hypothesized that there would be a reduction in the measured amplitude of the LV pacing stimulus on the skin surface using a high-resolution (HR) ECG when pacing from LV regions with scar compared with regions without scar. We term this the ECG Amplitude Signal Evaluation (EASE) method. METHODS: Consecutive patients with ischemic LV systolic dysfunction and standard criteria for de novo cardiac resynchronization therapy implantation were prospectively enrolled. All underwent a preimplant contrast-enhanced cardiac magnetic resonance study to assess for scar. The average amplitude of the LV pacing impulse was sampled on HR surface ECG intraprocedurally and then compared with the cardiac magnetic resonance results. RESULTS: A total of 38 LV pacing sites were assessed among 13 recipients. The median voltage measured on the surface HR ECG in regions with scar was reduced by 41% (interquartile range, 17% to 63%), whereas there was no measurable change in voltage (interquartile range, 0 to 0%) in regions without scar compared with the maximal amplitude (Wilcoxon P < 0.0001). CONCLUSION: The EASE method appears to be of potential value as a novel intraoperative tool to guide LV lead placement to regions free of scar. Future work is required to validate the utility of this method in a larger patient cohort.

CONTEXTE: La mise en place de la sonde ventriculaire gauche dans une zone exempte de cicatrice myocardique est un facteur déterminant de la réponse au traitement de resynchronisation cardiaque. Nous avons cherché à développer et à valider une approche électrocardiographique (ECG) simple, concrète et novatrice afin de repérer de manière peropératoire la présence de tissu cicatriciel au ventricule gauche (VG). Nous avons émis l'hypothèse qu'il y aurait une diminution de l'amplitude du rythme de stimulation du ventricule gauche mesurée à la surface de la peau à l'ECG haute résolution (HR) lors de la stimulation de régions du VG présentant du tissu cicatriciel comparativement aux régions exemptes de cicatrices. Il s'agit de ce que nous appelons la méthode EASE ( E CG A mplitude S ignal E valuation). MÉTHODOLOGIE: Des patients vus de manière consécutive qui présentaient une dysfonction systolique ischémique du VG et répondaient aux critères standard pour l'implantation de novo d'un dispositif de resynchronisation cardiaque ont été recrutés de manière prospective. Tous ont fait l'objet d'une résonance magnétique cardiaque améliorée par injection d'un produit de contraste avant l'implantation pour évaluer la présence de tissu cicatriciel. L'amplitude moyenne de l'impulsion de stimulation du VG a été échantillonnée sur l'ECG de surface HR réalisé pendant l'intervention, puis comparée aux résultats de la résonance magnétique cardiaque. RÉSULTATS: En tout, 38 points de stimulation du VG ont été évalués chez 13 receveurs. Le voltage médian mesuré sur l'ECG de surface HR dans les régions présentant du tissu cicatriciel était réduit de 41 % (intervalle interquartile : 17 % à 63 %), tandis qu'il n'y avait pas de changement mesurable du voltage (intervalle interquartile : 0 à 0 %) dans les régions exemptes de cicatrices par rapport à l'amplitude maximale (test de Wilcoxon, p < 0,0001). CONCLUSION: La méthode EASE semble avoir une utilité potentielle en tant que nouvel outil peropératoire pour guider la mise en place de la sonde ventriculaire gauche dans les régions exemptes de cicatrices. Il faudra réaliser d'autres travaux pour valider l'utilité de cette méthode dans une cohorte de patients plus importante.

Electrocardiographic morphology changes with different type of repolarization dispersions.

BACKGROUND: T-wave morphology changes have been linked to heterogeneity of ventricular repolarization and increase of arrhythmia vulnerability. Therefore, century-long debates around the genesis of T wave become even more relevant. Here are some interesting questions for the debates: (1) why T waves are usually concordant with QRS complex? (2) Is there a significant and consistent transmural dispersion of repolarization across heart wall? (3) What kind of T-wave morphology changes can be induced by either transmural or apical-basal dispersion of repolarization? METHOD: The previously developed GE's cell-to-electrocardiogram (ECG) model (GE Healthcare, Milwaukee, WI) was used to study the relation between cellular behavior and the T-wave morphology. The study focused on 2 types of repolarization dispersions: (1) Transmural (from endocardium to epicardium) and (2) Apical-basal (from apex to base of ventricles). More specifically, the transmural dispersions were created by adjusting the slow and fast delayed potassium rectifier current (Iks, Ikr) and transient outward current (Ito), on endocardial, midmyocardial (M cell) and epicardial cells separately. The apical-basal dispersion was adjusted according to the coordinates along the axis from the base to the apex of the ventricle. The contribution of M cell toward T-wave morphology were studied by adjusting the M cell's repolarization time in the range of shorter to longer than those of endocardial repolarization time. RESULTS: In the global transmural dispersion cases, QT interval is prolonged from 350 to 450 milliseconds, T-peak to T-end interval (TpTe) is prolonged from 50 to 130 milliseconds, and T-wave notches appeared when the heterogeneity is increased. In the localized transmural dispersion cases, significant T-wave morphology features such as TpTe, T-wave notches appeared in very limited precordial leads. In the global apical-basal dispersion cases, main T-wave change is on the amplitude, and T waves in several precordial leads and lead II turn to positive from negative. And the localized apical-basal dispersion does not generate significant T-wave morphology changes. CONCLUSIONS: The cell-to-ECG model provides a unique way to study electrophysiology and to link physiologic factors to ECG morphology changes. The simulation results suggest that the apical-basal dispersion of repolarization contributes to positive T wave more than the transmural dispersion. The contribution of localized transmural dispersion to surface ECG is very much localized to certain precordial leads.

Computerized STEMI recognition: an example of the art and science of building ECG algorithms.

With the advent of thrombolytics, guidelines for ST-elevated myocardial infarction (STEMI) recognition were presented in terms of an ST segment exceeding a particular level (1 or 2 mm) in 2 contiguous leads. However, more than half of prehospital electrocardiograms that exceed these ST criteria are from patients not having an acute myocardial infarction. In contrast, expert physicians (EXMD) maintain a high specificity (>95%) for the recognition of STEMI. Likewise, in terms of increasing sensitivity, it has been found that the EXMD will classify STEMI at lower levels than specified in the guideline. Thus, the EXMD uses additional electrocardiogram features to identify patients for appropriate intervention. Given that STEMI can be defined in terms of a pattern that is recognized by the EXMD as well as a clinical classification that can be evaluated in terms of clinical outcomes, the development and validation of a computer algorithm for STEMI need to include both the art of understanding how the human is detecting STEMI as well as the science required to develop quantified criteria based on clinical outcomes. Evidence is presented that demonstrates that reciprocal depression is a strong indicator of STEMI versus other causes of ST elevation.

Reference values of electrocardiogram repolarization variables in a healthy population.

INTRODUCTION: Reference values for T-wave morphology analysis and evaluation of the relationship with age, sex, and heart rate are lacking in the literature. In this study, we characterized T-wave morphology in a large sample of healthy individuals. METHOD: A total of 1081 healthy subjects (83% men; range, 17-81 years) were included. T-wave morphology variables describing the duration, area, slopes, amplitude, and distribution were calculated using 10-second digital electrocardiogram recordings. Multivariate regression was used to test for dependence of T-wave variables with the subject age, sex, and heart rate. RESULTS: Lead V5 (men vs women) T-wave variables were as follows: amplitude, 444 versus 317 muV; area, 48.4 versus 33.2 ms mV; Tpeak-Tend interval, 94 versus 92 milliseconds; maximal descending slope, -5.15 versus -3.69 muV/ms; skewness, -0.24 versus -0.22; and kurtosis, -0.36 versus -0.35. Tpeak-Tend interval, skewness, and kurtosis were independent of age, sex, and heart rate (r(2) < 0.05), whereas Bazett-corrected QT-interval was more dependent (r(2) = 0.40). CONCLUSION: A selection of T-wave morphology variables is found to be clinically independent of age, sex, and heart rate, including Tpeak-Tend interval, skewness, and kurtosis.

Robust QT interval estimation--from algorithm to validation.

BACKGROUND: This article presents an effort of measuring QT interval with automatic computerized algorithms. The aims of the algorithms are consistency as well as accuracy. Multilead and multibeat information from a given segment of ECG are used for more consistent QT interval measurement. METHODS: A representative beat is generated from selected segment of each lead, and then a composite beat is formed by the representative beats of all independent leads. The end result of the QT measure is so-called global QT measurement, which usually correlates with the longest QT interval in multiple leads. Individual lead QT interval was estimated by using the global measurement as a starting point, and then adapted to the signal of the particular lead and beat. In general, beat-by-beat QT measurement is more prone to noise, therefore less reliable than the global estimation. It is usually difficult to know if difference of beat-by-beat QT interval is due to true physiological change or noise fluctuation. RESULTS: The algorithm was tested independently by a clinical database. It is also tested against action potential duration (APD) generated by a Cell-to-ECG forward-modeling based simulation signals. The modeling approach provided an objective test for the QT estimation. The modeling approach allowed us to evaluate the QT measurement versus APD. The mean error between the algorithm and cardiologist QT intervals is 3.95 +/- 5.5 ms, based on the large clinical trial database consisting of 15,910 ECGs. The mean error between QT intervals and maximum APD is 17 +/- 2.4, and the correlation coefficient is 0.99. CONCLUSIONS: The global QT interval measurement method presented in this study shows very satisfactory results against the CSE database and a large clinical trial database. The modeling test approach used in this study provides an alternative "gold standard" for QT interval measurement.