PACAC: PYNQ Accelerated Cardiac Arrhythmia Classifier with secure transmission- A Deep Learning based Approach

Abstract

Electrocardiogram (ECG) signals are vital features to identify a healthy body; diagnosing cardiovascular diseases (CVDs) automatically using computer-aided tools has caught a significant attention in the current medical scenario. In recent times with the rapid growth of smart health-care system, IoT enabled edge devices make it possible for early diagnosis of diseases with resource constraint devices. PYNQ- a Python productivity on Xilinx platform, based hybrid CNN architecture has been proposed in this work for classifying arrhythmia in reference to AAMI (Association for the Advancement of Medical Instrumentation) EC57 standard. A comparative investigation is conducted on volume of trainable parameters of the architecture, and accuracy of ECG classification. A customized FPGA IP for the proposed hybrid 1-D CNN architecture has been generated using Vitis High Level Synthesis (HLS) tool that would be implemented on PYNQ-Z2 board. A lightweight cryptographic algorithm ASCON has been used in the proposed framework, wherein an authentication-based scheme for verifying an individual, via corresponding ECG signals is used, prior to sharing their data with various health-care entities furthermore enhancing information privacy.