An Artificial Neural Network model taking raw ECG input data achieved around 97% accuracy and was implemented on a Xilinx Zybo FPGA in 232 clock cycles.
Does an Artificial Neural Network model using raw ECG data implemented on an FPGA accurately classify arrhythmias?
An Artificial Neural Network model using raw ECG data implemented on an FPGA achieved approximately 97% accuracy for arrhythmia classification.
Cardiovascular diseases (CVDs) are one of the major causes of mortality around the world. Hence, regular monitoring of electrocardiogram (ECG) signals is crucial for early diagnosis and treatment. This leads to the ASIC/FPGA implementation of ECG classification. The currently suggested FPGA developments depend on statistical analysis of ECG signals to extract some features as the input for the classification network. However, feature extraction methods may cause some information loss. Therefore, an Artificial Neural Network (ANN) model that takes raw input data has been proposed in this work. The MIT-BIH arrhythmia dataset is used for the training and validation of the model. The proposed architecture consists of 2 hidden layers and an output layer. The training achieves around 97% accuracy. The network parameters (weights and biases) are extracted from the trained model as 32-bit floating-point numbers and converted into fixed-point numbers (8-bit) for efficient mapping to the FPGA. Then, the mathematical model of the feed-forward network was developed on Xilinx Zybo FPGA using Verilog HDL. The whole procedure is completed in 232 clock cycles.
Dal et al. (2022) studied Arrhythmia. Artificial Neural Network (ANN) model on Xilinx Zybo FPGA was evaluated on Classification accuracy. An Artificial Neural Network model taking raw ECG input data achieved around 97% accuracy and was implemented on a Xilinx Zybo FPGA in 232 clock cycles.