ECG Multi-Class Classification using Neural Network as Machine Learning Model

The main objective of this paper is to prepare a Clinical Decision Support System (CDSS) for a multi-class classification of ElectroCardioGram (ECG) signals into certain cardiac diseases. This CDSS is based on Artificial Neural Network (ANN) as a machine learning classifier and uses time scale input features. Fourty eight (48) ECG signals were selected from MIT-BIH arrhythmia database, of one minute recording. Unfortunately, among several types of learning algorithms for the ANN classifier, finding the appropriate one demands a comparative study. So, in this study, we have evaluated the impact of two learning algorithms, which are the LevenbergMarquardt (trainlm) and the Bayesian-Regularization (trainbr) on the proposed ANN performance. Consequently, we have achieved that trainbr reaches the most accurate result (93.8%), while trainlm generates the highest classification speed (0.582s). Subsequently, in order to assess the efficiency of this work, a second comparative study with related works, is done. Therefore, despite not being in the same working conditions, the obtained accuracy (93.8%) is considered acceptable.