Heart Rate Variability and Irregularity in Atrial Fibrillation

Sammanfattning

Atrial fibrillation (AF) is the most common cardiac arrhythmia worldwide, affecting an estimated 59.7 million people and imposing a growing burden on public healthcare systems. AF is characterized by a rapid and irregular heart rate. Specifically, understanding how rate-control therapies affect RR-interval variability and irregularity is essential for optimizing treatment. Accurate and reliable heart rate variability (HRV) analysis depends critically on robust QRS-detection and quality assurance of the resulting RR-interval series. This thesis investigates HRV and beat-to-beat irregularity in patients with permanent AF using 24-hour Holter ECG recordings from the RATAF II study, a randomized trial comparing rate control with diltiazem and metoprolol. Three interconnected objectives were addressed. First, the Pan Tompkins QRS-detection algorithm was evaluated by comparing its output against RR intervals extracted by clinical hospital software, revealing generally close agreement with minor timing discrepancies attributable to the 125 Hz sampling resolution. Second, two quality assurance approaches were developed: an intersection-based method combining both detectors, and a Bidirectional Long Short-Term Memory (BiLSTM) deep learning classifier trained to distinguish clean beats from noise. The BiLSTM achieved 97.0\% accuracy, 94.2\% sensitivity, and 99.8\% specificity on the validation set, though a tendency to reject beats affected by baseline wander was identified as a limitation. Lastly, HRV parameters were computed across five distinct RR interval pipelines and compared systematically. Results were remarkably consistent across all pipelines, suggesting that for long 24-hour recordings in this patient population, the choice of QRS-detector and quality assurance method has limited impact on HRV metrics. Across all pipelines, rate control medication reduced heart rate and increased variance-based HRV parameters, while entropy-based irregularity measures remained stable, indicating that rate control does not alter the underlying irregular nature of AF.