Fiducial segment averaging to improve cardiac time interval estimates.

To improve the accuracy of fiducial point estimates, a technique was developed and tested that uses the coherence of relatively small segments within the P-QRS-T complex from beat to beat. The methodology is applied to digitally stored electrocardiogram traces from various sources. It can be applied to single or multiple leads. First, the individual QRS complexes are identified with a matched filter technique. The trigger time points obtained are fine-adjusted by cross-correlating in turn each individual complex with the average of the remainder complexes and then shifting the complex till maximum correlation is achieved. The process is repeated till no improvement can be attained. Then the individual fiducial points are identified and a similar fine-adjustment is made to the individual points with a segment of approximately 60 ms around the fiducial point. This process is performed in turn for P-onset, QRS-onset and offset, peak of T wave and end of T wave. The values of PR-interval, QRS-duration and QT follow from these measurements. The validity of the result can be assessed by visual inspection and by examining the standard deviation of the measurements obtained. From the above intervals, it is generally accepted that QRS-duration is fairly constant over a short recording duration, whereas PR and QT are inherently heart-rate dependent. Often a linear regression of QT onto RR can be calculated with a significant correlation coefficient, indicating a physiological QT/RR relationship in these recordings of relative short duration. It would be impossible to show this last relationship if the individual QT changes had not been preserved by the technique described.