BPS2026 – Characterization of mechanisms involved in pre- and post-atrial fibrillation episodes

Atrial fibrillation (AF) triggering is influenced by the sympathetic, parasympathetic, and intrinsic sinoatrial node (SAN) systems, each contributing uniquely to heart rate variability (HRV). Characterizing HRV dynamics in the pre-AF and post-AF segments may enable early warning and provide mechanistic insights. We analyzed long (10–24 h) ECG recordings from three PhysioNet databases (LTDB, MIT-AF, SHDB). Inter-beat intervals were segmented into 5-min windows with a 2.5-min step, and seventeen HRV features were normalized per record. An EWMA-based change-detection pipeline was used to define a stable non-AF baseline, apply a record-specific dynamic threshold, search around each episode to detect meaningful changes, and trace their start and end to define pre- and post- episode segments. Summary statistics are reported as medians and interquartile ranges (IQRs) after outlier trimming. Pre-episode changes most frequently involve HTI, CVNN, SD2, LF, and SDNN, with CVNN showing the largest typical deviation from baseline. Median pre-episode durations for these features were 21–62 min, with wide distributions (IQR: 93–187 min). Post-episode changes were most prominent in HTI, SD2, LF, normality, and MeanNN (often decreasing while others increase); median post-episode durations were 77–178 min (IQR: 46–212 min). Concurrent non-AF rhythms often prolonged apparent pre- or post-segments. Across 106 records, 90% exhibited at least one predictive feature. Overall, the dynamics are database-, feature-, and episode-dependent and support patient-specific strategies for early AF prediction and recovery monitoring, indicating that AF initiation and termination reflect multisystem interactions among sympathetic, parasympathetic, and sinoatrial-node mechanisms.