Computer Detection of Atrial Flutter

Background: As part of ongoing research into methods for rhythm analysis in the Glasgow Program, a new algorithm has been developed for the detection of atrial flutter. Methods: The new algorithm uses two approaches to detect atrial waves. The first approach uses existing methods for “P” wave detection, searching various leads for patterns of positive and negative gradients satisfying amplitude and duration criteria. A threshold crossing technique is applied to the first difference of leads II and VI. Atrial regularity is measured using a scoring method that takes account of undetected waves. This measure of regularity is used, along with the estimated atrial rate, as criteria for atrial flutter. To develop the algorithm, 422 ECGs, including 49 cases of atrial flutter, were used as a training set. Optimization was achieved using a trial and error approach. A further 462 ECGs, including 75 atrial flutters, were used as a test set. Results: Using the new algorithm, a sensitivity of 72.0% was achieved, with a specificity of 98.4%. This compares favorably to existing flutter detection logic in the Glasgow Program, which had a sensitivity of 26.7% and specificity of 99.7% on the same difficult test set. The combination of the old and the new algorithms gave a sensitivity of 78.7% with a specificity of 98.1%. Conclusions: The new algorithm is successful in detecting a wide variety of cases of atrial flutter. However, further refinement is required to deal with certain types of atrial flutter, particularly at high ventricular rates. A.N.E. 2000;5(4):358–364