Evolutionary optimization of a wavelet classifier for the categorization of beat-to-beat variability signals

The beat-to-beat variation of the QRS and ST-T signal was assessed in healthy volunteers and in patients with malignant tachyarrhythmias using a novel wavelet based classifier designed by an evolutionary algorithm. High-resolution ECGs were recorded in 51 healthy volunteers and in 44 CHD patients with inducible sustained VT. QRS and ST-T variability was analyzed in 250 sinus beats. In each patient a variability signal was created from the standard deviation of corresponding data points of all beats. The complete variability signal was used. Analysis of the whole variability signal with the wavelet classifier results in an improved diagnostic ability of beat-to-beat variability analysis.