Precision Requirements for Electrocardiographic Measurements Computed Automatically

Frank-lead electrocardiograms from 115 male adults were processed through the Veterans Administration automated electrocardiographic analysis system. After analog-to-digital conversion, the bit precision for each record was changed four times from the initial 12 bits to 10, 8, 7, and 6 bits. Statistics for a single cardiac cycle and for averaged values over a 10-s period were obtained for amplitudes, durations, and time intervals for the five electrocardiographic records per patient. As compared to 12-bit precision, QRS spatial amplitude differences for a single beat ranged up to 7, 25, 77, and 160 ¿V for 10-, 8-, 7-, and 6-bit precision data, respectively. For QRS durations, these differences ranged up to 4, 8, 30, and 56 ms. For measurements averaged over a 10-s period, QRS amplitude differences ranged up to S, 20, 57, and 127 MV, and QRS duration differences ranged up to 2, 3, 6, and 31 ms for 10-, 8-, 7-, and 6-bit precision data. Wave recognition errors are probably responsible for many of these measurement changes. It is concluded that 1) unless wave recognition algorithms are tailored to fit lower precision data, precision should not be reduced below 8 bits, and 2) the use of measurements averaged over several seconds has a clear advantage over individual cardiac cycle measurements.