Learning Techniques of ECG Arrhythmia Classification: A Review

Predicting cardiac arrhythmia for early detection of various cardiac conditions using Electrocardiogram (ECG) is a critical research area that can help to prevent adverse heart problems or even death. Cardiac Arrhythmia is a life threatening disease which can be effectively detected by analyzing the abnormalities in the heartbeat. Automated classification of the abnormal heartbeats using patterns and incidence based on non-linear dynamics of RR interval is a challenge for developing new algorithms. Analyzing the morphological features and quasi periodic nature of ECG to recognize the presence and lack of arrhythmia needs a complete survey of the work that is in existence. Several works have implemented in the recent years to automatically detect and classify the ECG signals using traditional algorithms as well as advanced machine learning algorithms. In this paper, we have presented a survey on those algorithms and its segmentation methods, preprocessing techniques and its feature descriptions. Classification challenges and limitations of existing approaches for arrhythmia, have been reviewed and deliberated. Here, our intention is to focus on classification of arrhythmia with highly assured outcomes that can be employed in decision support systems used in diagnosis. Furthermore, it display those results and the possible scope of machine learning methods as the most preferred way to predict cardiovascular diseases (CVD) based on the analysis of cardiac arrhythmia using ECG is presented as the outcome of the survey. The paper address the need of prediction of life threatening diseases like CVD by suggesting suitable methods to have a positive impact on clinical practices and improves the outcome of medical diagnostics.