Towards Closed-Loop Neuromodulation: AI Prediction of Ventricular Tachyarrhythmias from Multimodal Biosignals in a Porcine Ischemia Model
Preprint
- 26 September 2025
- preprint
- Published by Cambridge University Press (CUP) in Cambridge Open Engage
Abstract
Myocardial ischemia hyperactivates cardiac sensory neurons, disrupting autonomic balance through excessive sympathetic activity and predisposing the heart to fatal ventricular tachyarrhythmias (VTs). Neuromodulation therapies, including spinal cord stimulation, can suppress ischemia-induced sympathoexcitation. However, open-loop neuromodulation techniques lack adaptive control, leading to suboptimal precision, reduced long-term efficacy, and potential side effects. Closed-loop systems require reliable biofeedback to anticipate autonomic dysfunction and arrhythmic risk. In this study, we evaluate whether epicardial mapping, surface ECG, and arterial blood pressure can provide robust biofeedback signals for predicting VT incidence. Anesthetized Yorkshire pigs (n=12) were subjected to 1 hour of left anterior descending coronary artery ischemia, during which we recorded epicardial electrograms, surface ECG, and blood pressure. To confirm the increased sympathoexcitation and arrhythmogenicity during LAD ischemia, we measured the activation recovery interval, a surrogate for action potential duration, and the dispersion of repolarization. Additionally, we assessed the arrhythmia incidence by identifying the VT episodes. An AI method was then used to assess whether VT could be predicted using 20-second data segments leading to VT. During LAD ischemia, we observed shortened ARIs, increased DORs, and elevated arrhythmia scores, confirming the presence of sympathoexcitation and arrhythmogenicity. The AI model demonstrated a sensitivity of 0.774 and a specificity of 0.770, with a positive predictive value of 0.632 and a negative predictive value of 0.870. This study suggests that combining epicardial electrograms, surface ECG, and blood pressure may provide reliable inputs for an AI-assisted closed-loop system capable of predicting VT onset and triggering neuromodulation therapy before arrhythmia develops.Keywords
All Related Versions
This publication has 27 references indexed in Scilit:
- From motion to emotion: exploring challenging behaviors in autism spectrum disorder through analysis of wearable physiology and movementPhysiological Measurement, 2025
- Spinal Cord Stimulation Reduces Ventricular Arrhythmias by Attenuating Reactive Gliosis and Activation of Spinal InterneuronsJACC: Clinical Electrophysiology, 2021
- The autonomic nervous system and cardiac arrhythmias: current concepts and emerging therapiesNature Reviews Cardiology, 2019
- Thoracic spinal cord and cervical vagosympathetic neuromodulation obtund nodose sensory transduction of myocardial ischemiaAutonomic Neuroscience, 2017
- Thoracic Spinal Cord Stimulation for Heart Failure as a Restorative Treatment (SCS HEART study): First-in-man experienceHeart Rhythm, 2015
- Group Invariant ScatteringCommunications on Pure and Applied Mathematics, 2012
- Assessment of autonomic control of the heart during transient myocardial ischemiaJournal of Electrocardiology, 2012
- Sympathetic neural activity and the pathogenesis of sudden cardiac deathHeart Rhythm, 2008
- Myocardial ischemia and ventricular fibrillation: Pathophysiology and clinical implicationsInternational Journal of Cardiology, 2007
- Sympathetic nerve sprouting, electrical remodeling and the mechanisms of sudden cardiac deathCardiovascular Research, 2001