Arrhythmia Detection from ECG Signal Using Long Short Term Memory Network Model

Arrhythmia refers to the irregular heartbeats that are observable in ECG data. Traditional diagnostic techniques often struggle with accuracy and speed due to their complex and variable nature. We have constructed various machine learning and deep learning models to overcome these constraints. This paper proposed a novel approach using a Long Short-Term Memory network to classify arrhythmic heartbeats into five classes: normal sinus rhythm, left bundle branch block, right bundle branch block, tachycardia, and premature ventricular contraction. We trained and validated the basic LSTM model using the MIT-BIH Arrhythmia database results. The results were truly remarkable, with an accuracy of 98.85% that far surpasses other advanced methodologies. Among those are CNN-based, SVM-based, and hybrid models. A comparative analysis reveals that our technique exhibits superior accuracy and robustness in detecting various types of arrhythmias, potentially serving as a valuable tool for doctors to quickly and accurately identify the type of arrhythmia.