| Heart diseases have been one of the leading death causes in modern societies, for years. At the same time, early prevention and risk reduction are considered as the most efficient way for treatment. Arrhythmia is one of these kind diseases that can be detected and predicted from ECG signals. Basically, the sign of arrhythmia can be reflected from a long term ECG from the electrocardiograph. As a result, a long-term ECG can be used to detect or predict the signs of possible arrhythmia. Consequently, it is important to analyze and classify different ECG signals to prevent cardiovascular disease. On the other hand, the amount of health data is gradually expanding. Therefore, modern, fast, accurate, and scalable algorithms are required to analyze the enormous amount of health data. For this reason, this research is focused on extracting useful features from long term ECG using distributed feature extraction method, and classifying them using a novel proposed hybrid distributed classification algorithm, FVFI, to find the signs of arrhythmia among all tested ECGs. For this purpose, each ECG is analyzed and divided into several sub-signals by recognizing the heartbeat patterns using a signal analyzing algorithm. Then, the classifiable dataset is formed by extracting different comparable features from each sub-signal. The proposed classification algorithm, which is the improved version of the other algorithm known as VFI, is applied on the dataset to find all different heart activity patterns. Finally, the outcome of the proposed classification algorithm is compared with the results of using two other classification algorithms, KNN and VFI on the same dataset. The ECG dataset which is utilized in this research contains 30 minutes of high resolution ECGs from 42 individuals. Briefly, data analyzing process is done for more than 27 million single data records. |
