Clinical Validation of an Algorithm for Automatic Detection of Atrial Fibrillation from Single Lead ECG

Abstract

Atrial fibrillation (AF) is the most common cardiac arrhythmia encountered in the clinical practice in the western countries. People with AF usually have a significantly increased risk of stroke. Clinically, AF is diagnosed by a surface electrocardiogram (ECG). AF is characterized by the absence of P-waves and by a rapid irregular ventricular rhythm. The algorithms for automatic detection of AF either rely on the absence of P-waves or are based on ventricular rhythm variability (RR variability). This work presents an automatic algorithm for AF real time detection based on the analysis of the RR series (ventricular interbeat intervals) and of the difference between successive RR intervals (\({\it \Delta}{\rm RR}\) intervals). Coefficient of variation of \({\it \Delta}{\rm RR}\) series and Shannon Entropy of RR series, computed over 5 minutes segments, are used to discriminate AF from normal sinus rhythm. A classifier based on the Mahalanobis distance is then used.

The proposed algorithm was clinically validated on 61 patients with an history/suspect of intermittent AF. The results obtained show that the algorithm can precisely detect AF episodes from a 5-minute, single lead ECG, whit a specificity of 97.9%, a sensitivity of 100%, and an accuracy of 98.4%. Its implementation on a microcontroller makes it suitable as a home-care device for the accurate detection of AF episodes.