An FPGA-Based Cloud System for Massive ECG Data Analysis

Xu Wang, Yongxin Zhu, Yajun Ha, Meikang Qiu, Tian Huang

Abstract: In this brief, we propose a stand-alone system-on-a-programmable-chip (SOPC)-based cloud system to accelerate massive electrocardiogram (ECG) data analysis. The proposed system tightly couples network I/O handling hardware to data processing pipelines in a single field-programmable gate array (FPGA), offloading both networking operations and ECG data analysis. In this system, we first propose a massive-sessions optimized TCP/IP hardware stack using a macropipeline architecture to accelerate network packet processing. Second, we propose a streaming architecture to accelerate ECG signal processing, including QRS detection, feature extraction, and classification. We verify our design on XC6VLX550T FPGA using real ECG data. Compared to commercial servers, our system shows up to 38× improvement in performance and 142× improvement in energy efficiency.