36 min
Circulation: Arrhythmia and Electrophysiology August 2020 Issue Circulation: Arrhythmia and Electrophysiology On the Beat
-
- Natural Sciences
Paul J. Wang:
Welcome to the monthly podcast! On the Beat for Circulation: Arrhythmia and Electrophysiology. I'm Dr. Paul Wang, Editor-in-Chief. With some of the key highlights from this month's issue.
In our first paper, Demilade Adedinsewo and associates assess the accuracy of an artificial intelligence-enabled electrocardiogram [AI-ECG] to identify patients presenting with dyspnea who have left ventricular LV systolic function (defined as LV ejection fraction ≤35%) in the emergency department [ED]. Patients were included if they had at least one standard 12-lead electrocardiogram [ECG] acquired on the date of the ED visit and an echocardiogram performed within 30 days of presentation. Patients with prior LV systolic dysfunction were excluded. A total of 1,606 patients were included. Meantime from ECG echocardiogram was one day. The AI-ECG algorithm identified LV systolic dysfunction with an area under the curve [AUC] of 0.89 and accuracy of 85.9%. Sensitivity was 74%, specificity 87%, negative predictive value 97%, and positive predictive value 40%. To identify an ejection fraction less than 50%, the AUC was 0.85, sensitivity 86%, sensitivity 63%, and specificity 91%. NT-proBNP alone with a cutoff greater than 800 identified LV systolic function with an AUC of 0.80 by comparison.
In our next paper, Mahmood Alhusseini and associates hypothesize that convolutional neural networks [CNN] may enable objective analysis of intracardiac activation in atrial fibrillation [AF]. They perform panoramic recording of bi-atrial electrical signals in AF and use the Hilbert-transform to produce 175,000 image grids in 35 patients labeled for a rotational activation by experts who showed consistency, but with variability (kappa [κ]=0.79). In each patient, ablation terminated atrial fibrillation. A CNN was developed and trained on 100,000 AF image grids validated on 25,000 grids, and then tested on a separate 50,000 grids. They found in a separate test cohort of 50,000 grids, CNN reproducibly classified AF image grids into those with or without rotational sites with 95.0% accuracy. This accuracy exceeded that of support vector machines, traditional linear discriminant, and k-nearest neighbor statistical analyses. To probe the CNN, they applied gradient weighted class activation mapping, which revealed that the decision logic closely mimicked rules used by experts (C statistic 0.96). The authors concluded that convolutional neural networks improve the classification of intercardiac AF maps compared to other analyses and agreed with expert evaluation.
In our next paper, Kenji Okubo and associates examined whether late potential LP, abolition and ventricular tachycardia [VT] non-inclusive ability predicted long-term outcomes in patients with non-ischemic cardiomyopathy [NICM] undergoing VT ablation. The total 403 patients with NICM (523 procedures) who underwent VT ablation from 2010 to 2016 were included. The underlying structural disease consists of dilated cardiomyopathy (DCM, 49%), arrhythmogenic right ventricular cardiomyopathy (ARVD 17%), postmyocarditis (14%), valvular heart disease (8%), congenital heart disease (2%), hypertrophic cardiomyopathy (2%), and others (5%). Epicardial access was performed in 57% of patients. At baseline, the LPs were present in 60% of patients, and a VT was either inducible or sustained/incessant in 85% of the cases. At the end of the procedure LP abolition was achieved in 79% of cases in VT noninducability in 80%. After a multivariate analysis, the combination of LP abolition and VT noninducibility was independently associated with free survival from VT (hazard ratio, 0.45, p = 0.0002) and cardiac death (hazard ratio 0.38, P = 0.005). The benefit of LP abolition of preventing the VT recurrence in ARVD and postmyocarditis appeared superior to that observed for DCM.
In our next paper, Domenico Corradi, Jeffrey Saffitz and associates hypothesize that structural molecular changes in atrial myocardium th
Paul J. Wang:
Welcome to the monthly podcast! On the Beat for Circulation: Arrhythmia and Electrophysiology. I'm Dr. Paul Wang, Editor-in-Chief. With some of the key highlights from this month's issue.
In our first paper, Demilade Adedinsewo and associates assess the accuracy of an artificial intelligence-enabled electrocardiogram [AI-ECG] to identify patients presenting with dyspnea who have left ventricular LV systolic function (defined as LV ejection fraction ≤35%) in the emergency department [ED]. Patients were included if they had at least one standard 12-lead electrocardiogram [ECG] acquired on the date of the ED visit and an echocardiogram performed within 30 days of presentation. Patients with prior LV systolic dysfunction were excluded. A total of 1,606 patients were included. Meantime from ECG echocardiogram was one day. The AI-ECG algorithm identified LV systolic dysfunction with an area under the curve [AUC] of 0.89 and accuracy of 85.9%. Sensitivity was 74%, specificity 87%, negative predictive value 97%, and positive predictive value 40%. To identify an ejection fraction less than 50%, the AUC was 0.85, sensitivity 86%, sensitivity 63%, and specificity 91%. NT-proBNP alone with a cutoff greater than 800 identified LV systolic function with an AUC of 0.80 by comparison.
In our next paper, Mahmood Alhusseini and associates hypothesize that convolutional neural networks [CNN] may enable objective analysis of intracardiac activation in atrial fibrillation [AF]. They perform panoramic recording of bi-atrial electrical signals in AF and use the Hilbert-transform to produce 175,000 image grids in 35 patients labeled for a rotational activation by experts who showed consistency, but with variability (kappa [κ]=0.79). In each patient, ablation terminated atrial fibrillation. A CNN was developed and trained on 100,000 AF image grids validated on 25,000 grids, and then tested on a separate 50,000 grids. They found in a separate test cohort of 50,000 grids, CNN reproducibly classified AF image grids into those with or without rotational sites with 95.0% accuracy. This accuracy exceeded that of support vector machines, traditional linear discriminant, and k-nearest neighbor statistical analyses. To probe the CNN, they applied gradient weighted class activation mapping, which revealed that the decision logic closely mimicked rules used by experts (C statistic 0.96). The authors concluded that convolutional neural networks improve the classification of intercardiac AF maps compared to other analyses and agreed with expert evaluation.
In our next paper, Kenji Okubo and associates examined whether late potential LP, abolition and ventricular tachycardia [VT] non-inclusive ability predicted long-term outcomes in patients with non-ischemic cardiomyopathy [NICM] undergoing VT ablation. The total 403 patients with NICM (523 procedures) who underwent VT ablation from 2010 to 2016 were included. The underlying structural disease consists of dilated cardiomyopathy (DCM, 49%), arrhythmogenic right ventricular cardiomyopathy (ARVD 17%), postmyocarditis (14%), valvular heart disease (8%), congenital heart disease (2%), hypertrophic cardiomyopathy (2%), and others (5%). Epicardial access was performed in 57% of patients. At baseline, the LPs were present in 60% of patients, and a VT was either inducible or sustained/incessant in 85% of the cases. At the end of the procedure LP abolition was achieved in 79% of cases in VT noninducability in 80%. After a multivariate analysis, the combination of LP abolition and VT noninducibility was independently associated with free survival from VT (hazard ratio, 0.45, p = 0.0002) and cardiac death (hazard ratio 0.38, P = 0.005). The benefit of LP abolition of preventing the VT recurrence in ARVD and postmyocarditis appeared superior to that observed for DCM.
In our next paper, Domenico Corradi, Jeffrey Saffitz and associates hypothesize that structural molecular changes in atrial myocardium th
36 min