Compelling new electrocardiographic markers for automatic diagnosis

BACKGROUND AND OBJECTIVE

The automatic diagnosis of heart diseases from the electrocardiogram (ECG) signal is crucial in clinical decision-making. However, the use of computer-based decision rules in clinical practice is still deficient, mainly due to their complexity and a lack of medical interpretation. The objetive of this research is to address these issues by providing valuable diagnostic rules that can be easily implemented in clinical practice. In this research, efficient diagnostic rules friendly in clinical practice are provided.

METHODS

In this paper, interesting parameters obtained from the ECG signals analysis are presented and two simple rules for automatic diagnosis of Bundle Branch Blocks are defined using new markers derived from the so-called FMM_ecg delineator. The main advantages of these markers are the good statistical properties and their clear interpretation in clinically meaningful terms.

RESULTS

High sensitivity and specificity values have been obtained using the proposed rules with data from more than 35,000 patients from well known benchmarking databases. In particular, to identify Complete Left Bundle Branch Blocks and differentiate this condition from subjects without heart diseases, sensitivity and specificity values ranging from 93% to 99% and from 96% to 99%, respectively. The new markers and the automatic diagnosis are easily available at https://fmmmodel.shinyapps.io/fmmEcg/, an app specifically developed for any given ECG signal.

CONCLUSIONS

The proposal is different from others in the literature and it is compelling for three main reasons. On the one hand, the markers have a concise electrocardiographic interpretation. On the other hand, the diagnosis rules have a very high accuracy. Finally, the markers can be provided by any device that registers the ECG signal and the automatic diagnosis is made straightforwardly, in contrast to the black-box and deep learning algorithms.

Selvester QRS Score Predicts Improvement of LVEF in Atrial Fibrillation Patients with Systolic Heart Failure

BACKGROUND

Left-ventricular systolic dysfunction (LVSD) comorbid with atrial fibrillation is reversible, but recovery is limited in a subset of patients. The Selvester QRS (S-QRS) score is an electrocardiogram-based assessment that reportedly reflects myocardial scar/damage. We evaluated the predictability of S-QRS score for the recovery of left-ventricular ejection fraction (LVEF) in persistent AF (PeAF) patients with LVSD undergoing catheter ablation (CA).

METHOD

CA was performed in 51 PeAF patients with reduced LVEF (<40%); S-QRS scores were measured after restoration of sinus rhythm. LVEF was re-evaluated at one year after CA; LVEF recovery was related to the S-QRS score.

RESULTS

The median [interquartile range] S-QRS score was 1 points [0-2]. LVEF increased from 32% [28-37] at baseline to 56% [49-57] at one year after CA. Thirty-seven patients achieved normalization of LVEF (≥50%, Group A); 14 patients did not (Group B). Group A had significantly lower S-QRS scores than Group B (0 point [0-2] vs. 2 points [2-3], p<0.05). In univariate/multivariate analyses, S-QRS score was an independent predictor of LVEF normalization. In the receiver operating characteristic curve, the cut-off value of S-QRS score was 2 points for prediction of the LVEF normalization (AUC = 0.75). Patients with low S-QRS score (<2 points) had greater LVEF improvement than those with high S-QRS score (≥2 points, ΔLVEF: 23% [17-28] vs. 17% [12-24], p<0.05).

CONCLUSION

S-QRS scoring non-invasively assesses the improvement of LVEF in PeAF patients with LVSD after CA. A high S-QRS score may indicate underlying myocardial scar/damage associated with unknown etiologies for LVSD other than PeAF. This article is protected by copyright. All rights reserved.

Reaction-diffusion informed approach to determine myocardial ischemia using stochastic in-silico ECGs and CNNs

Every year, nine million people die globally from ischemic heart disease (IHD). There are many methods of early detection of IHD which can help prevent death, but few are able to determine the configuration and severity of this disease. Our study aims to determine the severity and configuration of ischemic zones by implementing the reaction-diffusion analysis of cardiac excitation in a model of the left ventricle of the human heart. Initially, this model is applied to compute twenty thousand in-silico ECG signals with stochastic distribution of ischemic parameters. Furthermore, generated data is effectively (r²=0.85) implemented for training a one-dimensional convolutional neural network to determine the severity and configuration of ischemia using only two lead surface ECG. Our results readily demonstrate that using a minimally configured portable ECG system can be instrumental for monitoring IHD and allowing early tracking of acute ischemic events.

The ability of the electrocardiogram in left bundle branch block to detect myocardial scar determined by cardiovascular magnetic resonance

AIMS

We aimed to improve the electrocardiographic 2009 left bundle branch block (LBBB) Selvester QRS score (2009 LBSS) for scar assessment.

METHODS

We retrospectively identified 325 LBBB patients with available ECG and cardiovascular magnetic resonance imaging (CMR) with late gadolinium enhancement from four centers (142 [44%] with CMR scar). Forty-four semi-automatically measured ECG variables pre-selected based on the 2009 LBSS yielded one multivariable model for scar detection and another for scar quantification.

RESULTS

The 2009 LBSS achieved an area under the curve (AUC) of 0.60 (95% confidence interval 0.54-0.66) for scar detection, and R² = 0.04, p < 0.001, for scar quantification. Multivariable modeling improved scar detection to AUC 0.72 (0.66-0.77) and scar quantification to R² = 0.21, p < 0.001.

CONCLUSIONS

The 2009 LBSS detects and quantifies myocardial scar with poor accuracy. Improved models with extensive comparison of ECG and CMR had modest performance, indicating limited room for improvement of the 2009 LBSS.

The 43rd International Society for Computerized Electrocardiology ECG initiative for the automated detection of strict left bundle branch block

The presence of left bundle branch block (LBBB) is an important predictor of benefit from cardiac resynchronization therapy (CRT). New "strict" electrocardiographic (ECG) criteria for LBBB have been shown to better predict benefit from CRT. The "strict" LBBB criteria include: QRS duration ≥140 ms (men) or ≥130 ms (women), QS- or rS-configurations of the QRS complex in leads V1 and V2, and mid-QRS notching or slurring in ≥2 of leads V1, V2, V5, V6, I and aVL. The "strict" LBBB criteria are not regularly used and most hospital automated ECG systems and physicians still use more conventional LBBB criteria. As part of the 43rd International Society for Computerized Electrocardiology (ISCE) meeting, we conducted an initiative on the automated detection of "strict" LBBB where industry and academic investigators could present their algorithm results on digital 12-lead ECGs with varying QRS morphologies from the MADIT-CRT trial (300 training and 302 test set ECGs that were manually adjudicated for "strict" LBBB presence). The results revealed a 64-82% accuracy, 48-76% sensitivity and 46-87% specificity for automated "strict" LBBB detection from 7 participants. Most mismatches were likely attributed to differences in detection and absence of specific definitions for notches and slurs while differences in QRS duration and S-waves in leads V1 and V2 were less problematic. The full unblinded training and test datasets including all ECG signals are being made available through the Telemetric and Holter ECG Warehouse (THEW) for further exploration.

Semi-automated QRS score as a predictor of survival in CRT treated patients with strict left bundle branch block

BACKGROUND

Cardiac Resynchronization Therapy (CRT) is widely used for treating selected heart failure patients, but patients with myocardial scar respond worse to treatment. The Selvester QRS scoring system estimates myocardial scar burden using 12-lead ECG. This study's objective was to investigate the scores correlation to mortality in a CRT population.

METHODS AND RESULTS

Data on consecutive CRT patients was collected. 401 patients with LBBB and available ECG data were included in the study. QuAReSS software was used to perform Selvester scoring. Mean Selvester score was 6.4, corresponding to 19% scar burden. The endpoint was death or heart transplant; outcome was analyzed using Cox proportional hazards models. A Selvester score >8 was significantly associated with higher risk of the combined endpoint (HR 1.59, p=.014, CI 1.09-2.3).

CONCLUSION

Higher Selvester scores correlate to mortality in CRT patients with strict LBBB and might be of value in prognosticating survival.

Evaluation of the ECG based Selvester scoring method to estimate myocardial scar burden and predict clinical outcome in patients with left bundle branch block, with comparison to late gadolinium enhancement CMR imaging

BACKGROUND

Myocardial scar burden quantification is an emerging clinical parameter for risk stratification of sudden cardiac death and prediction of ventricular arrhythmias in patients with left ventricular dysfunction. We investigated the relationships among semiautomated Selvester score burden and late gadolinium enhancement-cardiovascular magnetic resonance (LGE-CMR) assessed scar burden and clinical outcome in patients with underlying heart failure, left bundle branch block (LBBB) and implantable cardioverter-defibrillator (ICD) treatment.

METHODS

Selvester QRS scoring was performed on all subjects with ischemic and nonischemic dilated cardiomyopathy at Skåne University Hospital Lund (2002-2013) who had undergone LGE-CMR and 12-lead ECG with strict LBBB pre-ICD implantation.

RESULTS

Sixty patients were included; 57% nonischemic dilated cardiomyopathy and 43% ischemic cardiomyopathy with mean left ventricular ejection fraction of 27.6% ± 11.7. All patients had scar by Selvester scoring. Sixty-two percent had scar by LGE-CMR (n = 37). The Spearman correlation coefficient for LGE-CMR and Selvester score derived scar was r = .35 (p = .007). In scar negative LGE-CMR, there was evidence of scar by Selvester scoring in all patients (range 3%-33%, median 15%). Fourteen patients (23%) had an event during the follow-up period; 11 (18%) deaths and six adequate therapies (10%). There was a moderate trend indicating that presence of scar increased the risk of clinical endpoints in the LGE-CMR analysis (p = .045).

CONCLUSION

There is a modest correlation between LGE-CMR and Selvester scoring verified myocardial scar. CMR based scar burden is correlated to clinical outcome, but Selvester scoring is not. The Selvester scoring algorithm needs to be further refined in order to be clinically relevant and reliable for detailed scar evaluation in patients with LBBB.

Validation of an automatic diagnosis of strict left bundle branch block criteria using 12-lead electrocardiograms

AIMS

Strict left bundle branch block (LBBB) criteria were recently proposed to identify LBBB patients to benefit most from cardiac resynchronization therapy (CRT). The aim of our study was to automate identification of strict LBBB in order to facilitate its broader application.

METHODS

We developed a series of algorithms to automatically detect and measure parameters required for strict LBBB criteria and proposed a definition of QRS notch detection. The algorithms were developed using training (n = 20) and validation (n = 592) sets consisting of signal-averaged 12-lead ECGs (1,000 Hz sampling) recorded from 612 LBBB patients from Multicenter Automatic Defibrillator Implantation Trial-CRT. Four trained clinicians independently performed adjudication on 148 different ECGs for comparing automatic and manually adjudicated results, in addition to 13 ECGs for evaluation of intraobserver variability and 32 ECGs for interobserver variability. We assessed the performance of the automated algorithms using manually adjudicated ECGs as references.

RESULTS

Overall sensitivity and specificity for detecting strict LBBB were 95% and 86%, respectively. The mean absolute deviation (MAD) of QRS duration and notch/slur locations for the automated method versus the manual method was below 1 ms, and MAD values were lower than 2 ms for interobserver and intraobserver variability. Sensitivity and specificity for detecting notch and slur locations were 87% and 96% for notches and 78% and 90% for slurs using the automatic method. In addition 95% and 93% agreements for notches and 90% and 88% agreements for slurs were reached for intra- and interobserver.

CONCLUSION

The proposed algorithms automatically measure QRS features for the diagnosis of strict LBBB. Our study shows good performance in reference to manual results.