Novel technique to average breathing loops for infant respiratory function testing.

Abstract

Breathing loops can be obtained by plotting two respiratory signals on an x-y diagram: the resulting loops represent a non-parametric description of the respiratory system. In infancy, loops are commonly measured during tidal breathing and their interpretation is hampered by high within-subject variability. Therefore a two-dimensional averaging technique for loops has been developed. The algorithm is based on segmentation of the loops and required two steps. First, the total length of the loop of every breathing cycle was divided into a specified number of equidistant intervals and the co-ordinates calculated by stepwise linear approximation of the curve. Second, averaged loops were calculated using the arithmetical mean (or the median if there were artifacts) of the x and y co-ordinates of the loops for all calculated points. To compare the new technique with averaging in the time domain a simulation study was performed using respiratory signals with a coefficient of variation (CV) of 5%, 10%, 15% and 20%. In contrast to the new technique, with increasing CV, averaging in the time domain led to increasing contortions in the averaged flow-volume loops. Mean errors of peak tidal expiratory flow were -3.3%, -13.9%, -21.3% and -34.2%, whereas errors with the new technique were considerably lower (0.5%, 0.4%, 0.5% and -0.1%) and independent of the level of CV.