ECG-Based Arrhythmia Detection SNN Algorithm Using STDP and Spike Inference for Smart Health City

Abstract

Due to changes in the medical system, an aging population, and the current COVID-19 crisis, a lot of interest and investment in the healthcare industry is increasing. The development of machine learning technology led by deep learning drives its applications to healthcare services. However, the conventional deep learning models consume a lot of power and are difficult to miniaturize, so they are not suitable for wearable devices required by smart cities. Accordingly, interests in spiking neural network (SNN), a next-generation artificial intelligence research field with low power and parallelism, are emerging as an alternative. In this paper, SNN is designed with STDP, an unsupervised learning algorithm that learns time difference between pre and post synapses. It is also applied to ECG arrhythmia data sets and produces accuracy, sensitivity and specificity according to cumulative spike-based inference. With the experiment results, we want to show the possibility of SNN as a healthcare technology and new implications too.

Wongyu Lee

Researcher, FOUNT

His research interests include machine learning algorithms, specifically spiking neural network (SNN) and reinforcement learning.

Seungwoo Han

PhD Student, Tokyo University of Agriculture and Technology

His research interests include graph signal processing and representation learning.

Cheolsoo Park

Professor

His research interests include machine learning, adaptive signal processing, computational neuroscience, and wearable technology.