Atrial fibrillation classification based on MLP networks by extracting Jitter and Shimmer parameters

Atrial fibrillation (AF) is the most common cardiac anomaly and one that potentially threatens human life. Due to its relation a variation in rhythm during indeterminate periods, long-term observations are necessary for diagnosis. With increase data volume, fatigue complexity of features make analysis an increasingly impractical process. Most medical diagnostic aid systems based on machine learning, designed automatically detect, classify or predict certain behaviors. In this work, using PhysioNet MIT-BIH Fibrillation database, system MLP artificial neural network proposed differentiate, between AF non-AF, segments ECG’s features, obtaining average accuracy 80.67% test set, 10-fold cross-validation method. As highlight, extraction jitter shimmer parameters from ECG windows presented compose input sets, indicating slight improvement model’s performance. Added these, Shannon’s logarithmic energy entropies determined, also performance related use fewer features.