Efficient detection of cardiac arrhythmias using low-latency fpga-accelerated hybrid deep learning models

Abstract This study proposes a hybrid CNN-LSTM model for FPGA-accelerated cardiac arrhythmia detection. Detecting cardiac arrhythmias is essential for early intervention, lowering heart disease risk. However, real-time detection necessitates efficient computing because ECG data processing is difficult. The suggested CNN-LSTM model is intended to satisfy the minimal latency and high-throughput requirements of portable medical devices. It is implemented on FPGA via Intel’s oneAPI. Without sacrificing accuracy, quantization and pipelining techniques increase model efficiency even further. The evaluation of key performance parameters, such as efficiency, delay, and power consumption, on FPGA, CPU, and GPU platforms revealed that FPGA had a 5x increase in power efficiency and a 3x decrease in latency. This study shows how FPGA deployment with oneAPI allows for energy-efficient, real-time arrhythmia detection, which makes it appropriate for continuous applications in healthcare.