Automated ECG diagnosis of atrial flutter by means of wavelet transform

The authors report a new algorithm for the automated diagnosis of atrial flutter (AF) by computer ECG programs. It consists of applying the wavelet transform (WT) technique to the detection of atrial flutter waves in the inferior lead Y of the Frank orthogonal EGG. The algorithm was developed and tested on a learning group of 325 cases with 47% having AF. Its clinical usefulness was prospectively assessed on a testing group of 1344 patients with 1.8% having AF. The sensitivity in this testing group was 44% It was raised to 80% by including the diagnostic statement of "possible AF", but at the same time the positive predictive value decreased from 84.6% to 66%. However, the specificity remained quite high: 99.2%. This WT procedure allowed the authors to improve the performance of the automated diagnosis of AF as compared with their previous experience.<<ETX>>