| Summary: | The death rate from heart disease, especially coronary heart disease and heart rhythm disorders classified as very high. Early detection and treatment of heart disease can prevent permanent damage to the heart tissue. The same ECG signal, which is obtained from an electrocardiograph, can be interpreted differently by doctors. It is caused due to a variety of heart disease very much so to properly diagnose certain heart disorders patients required special skills and adequate experience.
This research will be implemented identification of high-risk cardiac abnormalities using Artificial Neural Network (ANN). The type of heart disease that will be examined include atrial fibrillation, ventricular tachycardia, ventricular fibrillation, and Ischemic Heart Disease / Coronary heart disease. The first stage is a preprocessing research that includes segmentation, spatial data transformation into one dimension, elimination of noise, uniformity sampling frequency, feature extraction using wavelet decomposition and data normalization, while the next step is the identification of ECG signals using backpropagation neural network. These stages apply to the process of training and testing process. Training data in the form of real data and simulation data are taken from the MIT-BIH database. This study also displays a new short wave each cardiac conditions obtained by returning to the time domain of mean spectrum each heart condition.
The simulation results show the overall accuracy is 97%. The best accuracy (100%) was achieved in Atrial Fibrillation ECG, whereas the lowest accuracy (79%) while recognizing the ECG Ventricular Fibrillation. The mean short wave correlation results relevant to the ECG signal by 60,27%
Keywords-ANN Backpropagation, ECG, feature extraction, wavelet packet
|
|---|