Cardiac Arrhythmia Prediction Using Improved Multilayer Perceptron Neural Network

Electrocardiogram (ECG) has much diagnostic information to ensure proper clinical decisions in cardiac arrhythmia. Heart rate variations are signposts of current heart disease or imminent cardiac diseases.This study uses an ECG to determine bundle branch block (BBB), a form of heart block involving conduction delay/failure in the heart’s bundle branch. Machine learning and data mining methods are considered to improve ECG arrhythmia detection accuracy. Usually an automated classification of cardiac arrhythmias procedure is suggested on the basis of linear and non-linear HRV analysis. This paper presents an automated method for classification of cardiac arrhythmic based on ECG rhythm. RR intervals are extracted from ECG data using Symlet, and symmetric uncertainty is used for feature reduction. Extracted RR data is the classification feature with beats being classified through a Improved Neural Network and finally being evaluated through the use of MIT-BIH arrhythmia database.