Impact of Signal Preprocessing on the Inverse Localization of the Origin of Ventricular Pacing

The ECG signal measured in clinical or experimental conditions is usually contaminated by noise, which should be eliminated by proper signal processing methods. The selection of the method can be crucial for the results obtained from measured signal. Seven methods of high frequency noise removal and four methods of baseline drift removal were applied on experimental ECG signals measured during 30 s of induced ventricular pacing. Then for all possible combinations of the signal pre-processing methods an averaged signal representing one heart beat was computed. The origin of ventricular activity was assessed by the inverse solution using a single dipole. The impact of pre-processing on the quality of the inverse solution was observed. The localization error between the known position of the stimulating electrode and the computed origin of ventricular activation was more than 5 cm if the baseline of the processed signal did not coincide with the zero line. However, if the processed signal started from the zero value, an acceptable localization error from 0.7 to 1.3 cm was obtained. The inverse assessment of the origin of ventricular tachycardia using a single dipole is very sensitive to proper baseline drift removal. Because the inverse solution was computed for averaged signal, various methods of high frequency noise removal did not affect the results.