Prediction of Spontaneous Termination of Atrial Fibrillation with Supervised Neural Networks

This paper presents a proposal of ECG analysis, which determines a spontaneous termination of atrial fibrilation prediction. Supervised neural networks are trained to develop this task, where a comparison is carried out between multilayer perceptron (MLP) and supervised self organized maps (SOM). Principal component analysis (PCA) is implemented to reduce the input dimensionality. Results show maximum classification rates of 100% for MLP in the cases without and with PCA. For SOM the maximum classification rates are in 65% and 75% for case without and with PCA, respectively.