Heart Rate Variability Reveals Two Phenotypes of Atrial Fibrillation

Atrial fibrillation is the most prevalent type of heart arrhythmia. We suppose that patients with this heart condition can be classified by disease phenotypes. Previous phenotyping studies attempted classifications based on clinical information and genetic analysis. In this work, we derive two phenotypes of paroxysmal atrial fibrillation based on an extended set of heart rate variability properties using open access database of Holter ECG recordings of patients with paroxysmal and sustained atrial fibrillation. Analysis techniques used in this study include dimensionality reduction, use of multiple outlier detection methods and agglomerative clustering. Two derived phenotypes differ by 34 heart rate variability indices with high statistical significance. Specific indices suggest higher parasympathetic heart rhythm influence in one of the phenotypes. The major difference between the two phenotypes is the total duration of atrial fibrillation paroxysms in 24-hour ECG recordings.