Automatic detection of atrial fibrillation and flutter using the differentiated ECG signal

Atrial electrical activity during erratic heart rhythms, such as atrial fibrillation (AF) and flutter (AFL), are difficult to characterize quantitatively. In this work the authors investigated quantitative differences in ECG signals with AF and AFL. In each RR interval they analysed a set of windows to determine the characteristic waves of the AF and AFL (F waves). The method uses the differentiated and low-pass filtered ECG signal for the detection of F wave boundaries. The method has been applied to ECG records of the MIT-BIH database with a wide range of AF and AFL morphologies. The best statistical significance to separate AF from AFL has been obtained when analysing the variation coefficient of the F wave duration (VCFD): AF 0.368/spl plusmn/0.071, AFL 0.204/spl plusmn/0.068, p<0.001. When defining an ECG record as AF, if VCFD>0.255, the clinical performance of VCFD for detection of AF presents a sensitivity of 90%.