Optimizing the Automated Detection of Atrial Fibrillation Episodes in Long-term Recording Instrumentation

The aim of this paper was to optimize and evaluate the performance of our method for automated recognition of AF episodes that has been based on classification of the selected features derived from the heart rate signal. The aggregation of the classified heart beats has been added to minimize the false positive cases being noted by the clinicians when the wristband AF recorder was applied for long-term monitoring. Proposed improvement of the automated method led to considerable increase of sensitivity and improvement of the positive predictive value. At the same time the detection algorithm remains easy to be implemented in the mobile instrumentation for long-term monitoring.