Predicting Transthoracic Defibrillation Shocks Outcome in the Cardioversion of Atrial Fibrillation Employing Support Vector Machines

In this work, we use support vector machines (SVM) to predict if a defibrillation shock is likely to be successful or not in the cardioversion of persistent AF patients. The ECG signals of 47 patients elected for electrical cardioversion treatment were collected at the Royal Victoria Hospital in Belfast city, NI-UK. Signal processing was performed on ECG segments prior each shock. Three electrocardiographic indexes were extracted and used as input: the dominant atrial fibrillatory frequency, the mean and the standard deviation of the R-R interval time series of the ECG segments. We trained SVM using about 40% of the data. SVM could predict the outcome of 89% of low-energy shocks <= 100 [J], with a sensitivity (SE) of 87.50% and specificity (SP) of 98.8%. As a remarkable result, the outcome of higher energy shocks (>= 150 [J]) could be predicted with 100% exactitude.