BackgroundInformation about short Atrial Fibrillation (AF) episodes can be gathered from the diagnostic records of cardiac implantable electronic devices (CIEDs). CIEDs are not accurate when detecting short arrhythmia episodes. The correlation between mode switching events and AF episodes is significant for long events but prone to errors for short episodes. MethodsExpectation-maximization algorithms are used to estimate the parameters of a mathematical model from a list of AF episodes produced by the CIED. The durations of some of the episodes may be missing. Abnormal mode changes are detected and short episodes are joined into longer events when appropriate. The proposed method does not require that the sensitivity parameters of the device are altered. Post-processing of the data is limited to the detection of false negatives, thus paroxysmal arrhythmia diagnostic evaluations are safer. ResultsA three year-long study was carried out with patients with dual-chamber pacemakers (PM) at the Hospital Universitario Central de Asturias (Spain) between 2012 and 2015. The number of patients in which the proposed algorithm altered the final histogram was 40 out of 76. On average, the algorithm removes 2.79% of episodes shorter than 1 min in length and finds that 1% of the previously unaccounted episodes are longer than 30 min, of which 16% are longer than 24 h. ConclusionThe method is stable and guarantees that long arrhythmia episodes are never eliminated, and at the same time it is the most similar to the human expert in finding new long episodes.
Read full abstract