Work

Dynamic Risk Assessment of Severe Ventricular Arrhythmia in Patients With Implantable Defibrillators and Reduced Left Ventricular Ejection Fraction

Year: 2026
Type: article
Abstract: 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). RESULTS: <0.001). 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.
Language: English
Field: Medicine
Open Access status: hybrid
OpenAlex ID: W7237239648
FWCI:
Cites: 33
Cited by:
Related to: 10