Heart Abnormality Detection Using Acceleration Plethysmogram Signal

Abstract

For the past decade, irregularities of heart beat, or also known as Cardiac Arrhythmias has led to Cardiovascular Diseases (CVDs). It has been recorded as one of the main death cause in Malaysia as reported by the Ministry of Health Malaysia. This has become a major concern as many patients are unaware that early prevention can save them. In the past years, few systems that detects heart abnormality have been introduced. This includes various types of signal, such as the photoplethysmogram, electrocardiogram, electroencephalogram, ballistocardiogram and others. However, each of the systems have their own drawbacks. The study come out with a solution and proposed a system that can aid the problem, which is Heart Abnormality Detection Using Acceleration Plethysmogram (APG) signal. APG is more relevant as it displays obvious segments in the morphological cycle of the waveform. Throughout the study, datasets from PhysioNet, specifically MIMIC II Clinical Database waveform with sampling frequency of 125 Hz were used to obtain PPG and APG waveform. These signals undergoes five main processes before the abnormality can be detected. WEKA software was used for decision making process, where classifiers such as Multi-Layer Perceptron (MLP), Radial Basis Function (RBF) Network, and Bayesian Network were used to differentiate between the normal and abnormal heart beats. The percentage of correctly classified instances produced by the classifiers were able to show the significant differences between normal and abnormal APG signals. Based on the experimentation results, MLP and RBF classifier showed a significantly high classification accuracy of 96% as compared to Bayesian Network of 92%. This outcome suggest that APG signal can be used as an alternative in the detection of heart abnormalities with promising experimentation results.