Influence of time of sampling onset on parameters used for activation time determination in computerized...

The purpose of this work is to determine the sensitivity of the estimated time of peaks and maximum slopes, commonly used in activation time computations, to the instant at which sampling is initiated. Based on complex and quickly changing waveforms, 471 monopolar (MP) and bipolar (BP) epicardial responses in man were selected. These were decimated from 10 kHz to simulate sampling at frequencies ranging from 200 Hz to 2,000 Hz. The peak and maximum absolute slope for BP and the minimum slope for MP were computed repeatedly starting at successive 100 microseconds intervals extending throughout the sampling period and compared with these parameters computed from the waveform sampled at 10 kHz. Slopes were estimated using each of four different algorithms. The average greatest shift (AGS) due to variations in sampling onset ranged from 11.2 +/- 3.5 (200 Hz) to 0.3 +/- 0.2 msec (2,000 Hz). For bipolar algorithms, the peak performed better than the slope algorithms (AGS: 5.9 +/- 3.3 to 0.3 +/- 1.0 msec). For MP algorithms, 2 point linear, and 3 and 5 point Lagrange slope estimates performed similarly (AGS: 5.6 +/- 3.3 to 0.3 +/- 0.2 msec); a 5 point least square fit algorithm performed poorly. Sampling MP and BP electrograms below 500 and 400, respectively, often caused maximum shifts greater than 4 msec. Thus, the resolution of the peak and estimated slope is not limited to the sampling period, variations in initiation of sampling can cause significant outliers especially at low sampling rates, and MP electrograms should be sampled faster than BP electrograms for comparable accuracy.