Estudio comparativo de algoritmos para la detección de la fibrilación auricular

Abstract

Nine algorithms for Atrial Fibrillation (AF) detection were evaluated using the same protocol. The public databases MIT-BIH Arrhythmia and AF from PhysioNet were employed for the evaluation and comparison of these nine algorithms. Their performances were reported not only in terms of sensitivity (Se), specificity (Sp), positive predictive value (PPV), and error rate (Err), but also in terms of window length, input signal length, robustness to noise and computation time. These algorithms are based on the analysis of two characteristics observed in ECGs with AF: RR intervals Irregularity (RRI) and a chaotic atrial activity (AA). This AA can be analyzed in frequency (FA) and/or in time domain (P wave absence, PWA). Five algorithms were based on RRI; one algorithm was based only on AA; and the last three algorithms combined RRI with AA techniques.