Development of a Low-Cost ECG Device

This paper presents an approach to develop an affordable and easy-to-use electrocardiogram (ECG) monitoring device for early and timely detection of various cardiovascular diseases which would help in reducing the mortality rate of heart disease patients. Further, we present two different approaches to classify the signal into five different cardiac arrhythmia diseases as stated by the AAMI EC57 standard and compare their performance. The device comprises of an Arduino Uno and AD8232 heart rate monitor to record ECG and store it digitally on a computer. A three-lead based ECG electrode system is employed instead of the conventional twelve-lead based as it is cheaper and simpler to use. The preprocessing of the ECG signal such as filtering and QRS complex detection is done digitally using python. In the past few years, the neural network has established itself as a very powerful machine learning algorithm and has seen exponential growth in the field of biomedical signal and image analysis. Here, the performance and behavior of two different types of neural networks are evaluated, namely deep convolutional neural network (CNN) and feedforward neural network (FFNN) on the famous PhysioNet MIT-BIH Arrhythmia Database in classifying cardiac arrhythmia. Consequently, the prediction accuracy of 94.72 and 89.82% is achieved with CNN and FFNN respectively.