Algorithm for Identifying Patients with Paroxysmal Atrial Fibrillation without Appearance on the ECG

Although atrial fibrillation is the most common sustained cardiac rhythm disturbance, it remains underdiagnosed. One of the most drastic complications is embolism, and strokes in particular. Patients having atrial fibrillation must be identified in order to reduce the number of strokes. The algorithm presented detects atrial fibrillation, even without it being indicated in the analyzed ECG. Based on parameters of heart rate variability, only a 60-minute single channel ECG is required. At first, all R peaks are detected and all RR intervals are calculated. After normalizing the RR intervals, the time domain parameter SDSD is calculated and the so-called Poincare Plot is generated. The image and the time domain analysis assess a risk level, which determines whether the patient is suffering from atrial fibrillation. The resulting sensitivity calculated for ECG recordings from the MIT-BIH atrial fibrillation database is 91.5% and the specificity determined for the ECG recordings from the MIT-BIH normal sinus rhythm database is 96.9%. The sensitivity depends on the atrial fibrillation burden. Even if a burden of 0% is assumed, the results still prove satisfactory (sensitivity nearly 83%).