Influence of random noise on the accuracy of the indicator-dilution method.

Indicator-dilution curves (IDCS) represent nonnormalised probability density functions of indicator transit times from an injection to a sampling site. The reciprocal of the normalising constant (area A) and the first moment (mean transit time MTT) are then the basic variables for the estimation of cardiac output and volumes in circulatory studies. There have been numerous investigations on extrapolating the primary curve from the part that is not disturbed by recirculation, and also on the basic assumptions required for accurate measurements. An approach which is often used assumes single exponential decay leading to log-linear extrapolation into the tail of the curve. Recirculation, however, may cause this exponential to be wrongly estimated leading to overestimation of A and MTT. The authors objective was to investigate possible systematic errors occurring with a log-linear extrapolation when random noise was superimposed on the curve. This is particularly relevant for possible errors made by some so-called 'cardiac output computers' when there are fluctuations from an alternating flow fluctuations by Poisson noise or random sampling errors. The A and MTT estimates have been compared with those obtained by a nonlinear least-squares fit to a diffusion with drift model.