Hybrid Neuro-Fractal Analysis of ECG Signal to Predict Ischemia

In this paper, we proposed a new hybrid model to predict the number of ischemia occurrences in heart patients based on their ECG records using fractal analysis, statistical analysis, and Artificial Neural Networks (ANN). Fractal analysis is employed to determine the long-term correlations on ECG signal. The statistical analysis methods like binary logistic regression and multivariate linear regression are applied to investigate the correlation of parameters to increase model’s accuracy. Finally, selected features are fed into the feedforward neural networks to predict the number of ischemia occurrences. One of the advantages of this work compared to similar existing works is its capability to extract useful parameters from fractal analysis of the signals and the clinical characteristics to make a prediction. Our results compare well with those of the earlier works and the augmentation of the above-mentioned approaches proves the prediction accuracy.