Bandwidth-induced errors in parameters used for automated activation time determination during computerized intraoperative cardiac mapping: Theoretical limits
Image Processing, Computer-Assisted
Two parameters commonly used when determining the time of local activation during computerized intraoperative cardiac mapping are the time of the peak in bipolar electrograms (BP) and the time of the largest negative slope (LNS) in monopolar electrograms (MP). The dependence of these parameters upon bandwidth was studied. A database of fast MP and BP was compiled from intraoperative recordings collected from epicardial sock arrays in man. Based on 95%, 99% and 99.9% of total signal power, the bandwidths of each response were determined. After removing specified frequencies using the Fourier transform, errors in determining these parameters were computed and characterized with respect to mean square error (MSE), percent power bandwidth and frequency content. The Fourier transform was used to abruptly filter out undesired upper frequencies. Filtering BP at 200 Hz produced a 1-ms shift in the peak in 17.0% of the electrograms, 0.4% of the peaks shifted by 2 ms and none shifted by more than 2 ms. Comparable shifts in the LNS occurred when filtering MP at 400 Hz. Retention of 95% of the power in a BP rarely (0.9%) changed the time of its peak by more than 1 ms. For MP, retention of 99.9% of the power was required for similar performance (3.6%). Thus, although BP have greater signal power in the higher frequencies than MP, a greater bandwidth is required when recording monopolarly in order to preserve the small amount of power necessary to define the time of the LNS with accuracy comparable to determination of the peak in BP.