Discriminating Drug-Induced Ventricular Fibrillation via Wavelet Transforms and Deep Temporal Models


Abstract:

Ventricular fibrillation (VF) is a life-threatening arrhythmia whose electrophysiological characteristics vary under different pharmacological interventions. We propose a deep learning framework that combines wavelet-based representations with time-aware modeling to classify VF episodes induced by amiodarone, diltiazem, flecainide or no drug. Using electrocardiogram recordings from anesthetized dogs, we extracted features via Continuous (CWT) and Scattering (SWT) wavelet transforms, and trained long-short-term memory networks to capture temporal dynamics. Classification performance is evaluated across the VF segments and latent representations are visualized. The results show improved separability over time, with SWT models outperforming CWT and achieving up to the 68% macro F1 score in an independent test set. This score reflects physiological similarities between drug effects, particularly the consistent overlap between amiodarone and flecainide in classification and latent space. These findings highlight the value of interpretable and time-aware models for VF characterization analysis.

Año de publicación:

2025

Keywords:

    Fuente:

    scopusscopus

    Tipo de documento:

    Other

    Estado:

    Acceso restringido

    Áreas de conocimiento:

    • Sistema circulatorio
    • Aprendizaje profundo
    • Medicina interna

    Áreas temáticas de Dewey:

    • Farmacología y terapéutica
    • Enfermedades
    • Métodos informáticos especiales
    Procesado con IAProcesado con IA

    Objetivos de Desarrollo Sostenible:

    • ODS 3: Salud y bienestar
    • ODS 12: Producción y consumo responsables
    • ODS 17: Alianzas para lograr los objetivos
    Procesado con IAProcesado con IA