A Rigorous Analysis of Biomedical Edge Computing: An Arrhythmia Classification Use-Case Leveraging Deep Learning

Biomedical Edge computing is an exciting area of interdisciplinary research involving the Internet of Medical Things (IoMT) sensors and devices with lightweight Artificial Intelligence (AI) logic. To address the rapidly growing need for smart and portable biomedical devices with localized decision-making capability, we present a proof-of-concept logic-in-sensor design with an arrhythmia analytics use-case. Existing signal processing techniques for arrhythmia analytics such as discrete wave transform (DWT) and non-linear delay differential equation (DDE) lead to high complexity and computational burden on biomedical edge devices due to expensive preprocessing steps. As a solution, we propose a deep learning-based lightweight arrhythmia classification method leveraging a customized one-dimensional (1-D) convolutional neural network (CNN). A rigorous analysis of the proposed method's performances and generalization potential are assessed using four publicly available datasets.