Enabling On-Device Model Personalization for Ventricular Arrhythmias Detection by Generative Adversarial Networks
Hosted in Virtual Platform
Embedded System Design Methodologies
DescriptionImplantable Cardioverter Defibrillator (ICD) is an ultra-low-power device delivering in-time defibrillation on ventricular arrhythmias (VAs). The parameters of VAs detection mechanism on each recipient's ICD should be fine-tuned to achieve accurate detection. However, the process relies on the expertise and must be conducted manually by cardiologists. In this paper, we introduce a novel self-supervised on-device personalization of convolutional neural networks (CNNs) for VAs detection. We first propose a framework to enable on-device CNN update. Furthermore, we propose a generative model that learns to synthesize patient-specific intracardiac EGMs signals, which can be used as training data to improve patient-specific VAs detection.