Activation in unipolar cardiac electrograms: a frequency analysis

We have developed and tested several detectors of local activations in unipolar cardiac electrograms; the detectors are based on the frequency content of the waveforms. For this study, myocardial regions with no local electrical activity were created with cryoablation in canine ventricles, so that the characteristics of electrograms reflecting local activation could be compared with those with only distant electrical activity. For each electrogram, representations of the original signal were created using the output of bandpass filters; for each representation, the value of the maximum amplitude was taken as a measurement of the frequency content of the electrogram in that frequency band. The content of each frequency band and the first derivative of the signal were tested as discriminators between local and distant electrical activity. Combinations of frequency bands were also tested using a logistic regression technique; certain combinations provided better detection than any of the individual frequencies or the first derivative. The inclusion of frequencies between 500 and 1000 Hz improved the detection performance, suggesting that sampling rates of 1000 samples per second or less may not be adequate for optimal discrimination. A detector based on multivariate analysis of different frequency components of a signal may be more effective than single-band filtering in discriminating between local and distant electrical activity in the heart, especially when those components have very different magnitudes.