Comparison of adaptive and nonadaptive filters for reduction of power line interference in the ECG

We have investigated the relative performance of an adaptive and nonadaptive 60-Hz notch filter applied to an ECG signal. We evaluated the performance of the two implementations with respect to adaptation rate (or transient response time), signal distortion, and implementation complexity. We also investigated the relative effect of adaptive and nonadaptive 60-Hz filtering on ECG data compression. With a 360 Hz sample rate and an adaptation time of approximately 0.3 s for a 1 mV 60-Hz signal, the adaptive implementation is less complex and introduces less noise, particularly in the ST-segment, into a typical ECG signal. When applied to ECG signals, prior to data compression by average beat subtraction and residual differencing, the residual signal resulting from the adaptively filtered signal had an average entropy 0.37 bits per sample (bps) lower than the unfiltered signal. The nonadaptive 60-Hz filter produced an average entropy decrease of 0.08 bps relative to the unfiltered EGG.