A framework for detecting high-performance cardiac arrhythmias using deep inference engine on FPGA and higher-order spectral distribution

Abstract

Cardiac arrhythmias (CA) are critical health conditions. In such a case, highly accurate detection leads to better management and, therefore, better treatment. Here, this paper presents a novel high-performance detection framework for cardiac arrhythmias based on advanced signal processing algorithms with deep learning on a Field Programmable Gate Array (FPGA) towards achieving real-time performances and even higher accuracy. ECG signals are initially analyzed by using compressive sensing theory to obtain sparsity, and from that, the adaptive compressive sensing framework is created. This compressive sensing framework adapts the sensing matrix step by step through compression via the Hybrid Reptile Search Algorithm integrated with the Garra Rufa Algorithm (Hyb-RSA-GRA). The adapted sensing matrix renders signal reconstruction efficient through Bayesian Regularization-Backpropagation Neural Network (BRBNN). The new arrhythmia detection framework employs the possibility of higher-order spectral distribution (HoSD) in extracting finer patterns from ECGs that describe arrhythmia. The task of classification uses a pre-trained Graph Convolutional Neural Network (GCNN) acting as a Deep Inference Engine on the FPGA to support real-time, robust identification of the type of arrhythmias such as N (normal beat), S (supraventricular ectopic beat), V (ventricular ectopic beat), F (fusion beat), and U (unidentified beat). The proposed FPGA implementation reveals better performance with high accuracy, sensitivity, specificity, precision, recall, and F1-score with optimized power dissipation, resource utilization, and delay metrics. Furthermore, the compressive sensing framework guarantees low MSE, reduced RMSE, high SNR, and an improved reconstruction probability. All the above results demonstrate the capability of the framework in accurate prediction and hardware efficiency, hence making it a robust solution for cardiac arrhythmia detection.