Does a classification method using fractal dimension and BPN accurately identify 7 arrhythmias from ECG signals?
A novel technique using fractal dimensions and a Back Propagation Neural Network achieved 98.83% accuracy in classifying 7 arrhythmias from ECG signals.
The statistics indicate a dramatic increase in mortality due to cardiovascular failures in the worldwide. In developing countries, with lower per capita income, especially in rural areas, late diagnosis of the disease leads to the sudden death of people. Since electrocardiogram (ECG) is one of the most important tools for diagnosing cardiovascular diseases, this study has been presented to analyze this highly inexpensive and available signal. Extracted characteristics of the signal are a good representation of the heart function because of the chaotic, dynamic, and non-linear behavior of the heart. The fractal dimension is the best representative of the ECG signal which is able to take into account its hidden complexity. ECG signals are analyzed based on the fractal dimension and Back Propagation Neural Network (BPN). In this paper, a new technique is introduced for honest classification of 7 arrhythmias from ECG signals using the fractal dimension. This method is able to identify the exact location of arrhythmias. A combination of 5 reputable universal databases is used to classify based on the fractal dimension and BPN. The performance of this method is measured by Sensitivity (SE) and Specificity (SP) indices. According to the results, the accuracy of this method is equal to 98.83%.
Kiani et al. (Tue,) studied this question.