Abstract 2168: Efficient Detection of Atrial Fibrillation and Congestive Heart Failure in Single-Lead Devices

Abstract

Patients receiving primary prevention single lead ICDs are at risk for atrial fibrillation (AF) and congestive heart failure (CHF). No such device reports AF burden, and only a single CHF measure, trans-thoracic impedance, is available. Entropy measures that count the number of matching RR intervals have promise, as AF is random (high entropy) and CHF is often marked by reduced heart rate variability (RR intervals with many matches) and ectopic beats (few matches). We designed entropy-based measures to detect AF (high entropy) and CHF (mixture of RR intervals with many and with few matches). For real-world implementation, we used only 12 RR intervals, and calculated the result every 30 minutes in 24-hour Holter monitor records from the MIT-BIH databases. The Figure shows distinction among AF, NSR and CHF records using HR and S.D. (panel A) or the new entropy-based measures. Panel A shows poor diagnostic performance of conventional measures. In Panel B, the y-axis, COSEn, is the coefficient of sample entropy. The AF records all have higher values, and the ROC area is 1.00. The x-axis is a measure of template match counts. It distinguishes between normals and CHF patients with ROC area 0.92. With only 12 RR intervals every 30 minutes, entropy calculations allow for efficient detection of AF and CHF. We propose that single lead devices can be employed as monitors in the primary prevention population, where risk of AF and CHF is high.