optimization of ECG Peaks for Cardiac Abnormality Detection using Multilayer Perceptron

Artificial neural networks (ANNs) have are design based on computer adaptations to human biology (the concept of neurons). Extensive studies have conducted on the suitability of ANNs in applying various types of problems, especially in biomedical engineering. Applications to ANNs widely used for educational research as well as medical decision-making. In the current study, ANNs trained to perform the detection of cardiac abnormalities based on several reference parameters that are feed to the ANN. Reference parameters are better known as input parameters for cardiac abnormalities are the amplitude and duration extracted from Electrocardiogram (ECG) signals. An ECG complex contains a P peak, a QRS wave, and a T peak. Measurement on amplitude and measurement on duration of each P peak, QRS wave, and T peak are used as the input parameter, making six input parameters for the artificial neural network. In this work, a multilayer perceptron (MLP) is select as ANN structure along with three different training algorithms, which are Bayesian Regularization (BR), Lavenberg Marquardt (LM), and Backpropagation (BP). The MLP structure activated by the Tansig activation function. From the result, the MLP network trained by the BR training algorithm activated by Tansig activation function performed the highest with 94.04% of accuracy.