Identification of patients with atrial fibrillation using HRV parameters

• Published in: Biomedizinische Technik Vol. 53; no. 1; p. 8
• Main Authors: Kikillus, Nicole, Hammer, Gerd, Bolz, Armin
• Format: Journal Article
• Language: German
• Published: Germany 01.02.2008
• Subjects: Algorithms; Artificial Intelligence; Atrial Fibrillation - diagnosis; Atrial Fibrillation - physiopathology; Diagnosis, Computer-Assisted - methods; Heart Rate; Humans; Pattern Recognition, Automated - methods; Reproducibility of Results; Risk Assessment - methods; Risk Factors; Sensitivity and Specificity
• ISSN: 0013-5585
• DOI: 10.1515/BMT.2008.002

Atrial fibrillation is the most common sustained cardiac rhythm disturbance. One of the most drastic complications is embolism, particularly stroke. Patients with atrial fibrillation have to be identified. This can lead to early therapy and thus avoiding strokes. The algorithm presented here detects atrial fibrillation securely and reliably. It is based on a single-channel ECG, which takes 60 min. First, the R-peaks are detected from the ECG and the RR interval is calculated. To be independent from pulse variations, the RR interval is normalized to 60 bpm. A parameter of heart rate variability is calculated in time domain (SDSD) and the so-called Poincaré plot is generated. The image analysis of the figures of the Poincaré plot is made automatically. The results from analysis in time domain, as well as image analysis, yield a risk level, which indicates the probability for the occurrence of atrial fibrillation. Even if there is no atrial fibrillation in the ECG while analyzing, it is possible to identify patients with atrial fibrillation. The sensitivity depends on the burden of atrial fibrillation. Even if a burden of 0% is assumed, the results still prove satisfactory (sensitivity of nearly 83%).