Resumen:
The aim of this work is to use biomarkers extracted from high-resolution voltage maps of atrial fibrillation (AF) patients in order to make predictions about future "redo" procedures. We collected maps of the left atrium of 122 patients, prior of being treated for AF The bipolar voltage maps were extracted with the Rythmia system from Boston Scientific and subsequently analyzed in the MATLAB environment. The present study focuses on three biomarkers extracted from those maps. Two are associated with the bipolar voltage measurements on the map, i.e., the mean voltage and the voltage dispersion on the map. The third indicator is the area of the atrium evaluated from the map. The data are used for feeding a supervised classification algorithm. The output variable is a binary variable that is set to 1 if the patient will need a "redo" procedure in the twelve months following the cardiac intervention and 0 otherwise. We show that the biomarkers have some statistical power in predicting future outcomes. Especially the mean voltage on the map is the best predictor of the future outcome. We determine the cutoff value for the mean voltage based on the best prediction accuracy of Vm=0.542 mV in agreement with previous studies. We discuss some extensions of this study that could allow improvements in predictive power.
