FPGA-based system for artificial neural network arrhythmia classification

• Published in: Neural computing & applications 2020-04, Vol.32 (8), p.4105-4120
• Main Authors: Zairi, Hadjer, Kedir Talha, Malika, Meddah, Karim, Ould Slimane, Saliha
• Format: Article
• Language: English
• Subjects: Arrhythmia; Artificial Intelligence; Artificial neural networks; Cardiac arrhythmia; Classification; Computational Biology/Bioinformatics; Computational Science and Engineering; Computer Science; Data Mining and Knowledge Discovery; Diagnostic software; Diagnostic systems; Electrocardiography; Energy conservation; Energy consumption; Field programmable gate arrays; Floating point arithmetic; Image Processing and Computer Vision; Monitoring; Neural networks; Original Article; Performance evaluation; Probability and Statistics in Computer Science; Real time; Telemedicine
• ISSN: 0941-0643; 1433-3058
• DOI: 10.1007/s00521-019-04081-4

The automatic detection and cardiac classification are essential tasks for real-time cardiac diseases diagnosis. In this context, this paper describes a field programmable gates array (FPGA) implementation of arrhythmia recognition system, based on artificial neural network. Firstly, we have developed an optimized software-based medical diagnostic approach, capable of defining the best electrocardiogram (ECG) signal classes. The main advantage of this approach is the significant features minimization, compared to the existing researches, which leads to minimize the FPGA prototype size and saving energy consumption. Secondly, to provide a continuous and mobile arrhythmia monitoring system for patients, we have performed a hardware implementation. The FPGA has been referred due to their easy testing and quick implementation. The optimized approach implementation has been designed on the Nexys4 Artix7 evaluation kit using the Xilinx System Generator for DSP. In order to evaluate the performance of our proposal system, the classification performances of proposed FPGA fixed point have been compared to those obtained from the MATLAB floating point. The proposed architecture is validated on FPGA to be a customized mobile ECG classifier for long-term real-time monitoring of patients.