Key result
Dynamic CNN analysis of electrograms predicts polymorphic VT/VF with ~0.73 AUROC, outperforming static risk models.
Why the study?
Device shocks are associated with an adverse prognosis in patients with implantable cardioverter-defibrillators, motivating a dynamic predictive model using intracardiac electrogram changes to assess polymorphic ventricular tachycardia or VF risk in HFrEF.
Does a convolutional neural network model analyzing dynamic changes in intracardiac electrograms improve the prediction of first polymorphic ventricular tachycardia or ventricular fibrillation events compared to static risk assessment based on clinical variables in patients with implantable cardioverter-defibrillators and HFrEF?
Cohort (n=7,129)
Yes
Does a convolutional neural network model analyzing dynamic changes in intracardiac electrograms improve the prediction of first polymorphic ventricular tachycardia or ventricular fibrillation events compared to static risk assessment based on clinical variables in patients with implantable cardioverter-defibrillators and HFrEF?
A convolutional neural network model analyzing dynamic intracardiac electrogram changes (R-wave amplitude, QRS duration, and PVC burden) significantly outperforms static clinical risk assessment in predicting severe ventricular arrhythmias (VF or polymorphic VT) in HFrEF patients with ICDs.
Answers are grounded in this paper and cite their sources.
May support dynamic arrhythmia risk monitoring in reduced-EF ICD patients; leaves open whether it improves outcomes over static assessment.
BACKGROUND: Device shocks are associated with an adverse prognosis in patients with implantable cardioverter-defibrillators. This study aimed to develop and validate a predictive model based on dynamic changes in intracardiac electrograms to assess the risk of polymorphic ventricular tachycardia or ventricular fibrillation (VF) events in patients with implantable cardioverter-defibrillators and heart failure with reduced ejection fraction (HFrEF). METHODS: Experimental studies were conducted in pigs with HFrEF (N=8) to characterize intracardiac electrogram and surface ECG changes during HF decompensation. Sham-operated animals (N=6) were used as controls. In patients, the study used data from 2 independent cohorts of patients with ICDs. The first cohort included 5089 patients from 50 centers (199 218 tracings), leveraging electrical parameters associated with HFrEF decompensation in the experimental model to train and internally validate a dynamic convolutional neural network model for polymorphic ventricular tachycardia/VF prediction. External validation was performed in a second cohort of 2040 patients from 7 centers (369 845 tracings). In a clinical setting, a convolutional neural network-derived model predicted first VF or polymorphic ventricular tachycardia events by analyzing temporal patterns of R-wave amplitude, QRS duration, and premature ventricular complex burden. This approach yielded areas under the receiver operating characteristic curve of 0.787 (0.753-0.817) for the training cohort and 0.725 (0.709-0.739) for the external validation cohort. Dynamic risk prediction using the convolutional neural network model significantly outperformed static risk assessment based on clinical variables (area under the receiver operating characteristic curve, 0.617 [0.602-0.632]). CONCLUSIONS: A convolutional neural network-derived model, grounded in the pathophysiology of HFrEF, provides dynamic risk assessment of VF or polymorphic ventricular tachycardia events in patients with implantable cardioverter-defibrillator devices.
No takes yet. Share an insight, caveat, or question.
Diz-Díaz et al. (2026) conducted a cohort in Heart failure with reduced ejection fraction (HFrEF) with implantable cardioverter-defibrillators (n=7,129). Dynamic convolutional neural network model analyzing intracardiac electrograms (R-wave amplitude, QRS duration, PVC burden) vs. Static risk assessment based on clinical variables was evaluated on First ventricular fibrillation (VF) or polymorphic ventricular tachycardia events (AUROC 0.725 (external validation) vs 0.617 (static clinical model), 95% CI 0.709-0.739, p=<0.001). Dynamic CNN risk modeling of intracardiac electrograms predicted polymorphic VT/VF, outperforming static clinical assessment.
Synapse has enriched 10 related papers. Consider them for comparative context: