Stewart-Hamilton and gamma variate methods in computer analysis of dye curves

The conventional Stewart-Hamilton (SH) principle was compared to the gamma variate function fitting (G) method in on-line computer determination of cardiac output from dye dilution curves on anaesthetized animals (dogs, cats and reindeer), and on an artificial circulation model with known flows. In animal experiments, indicator was injected into the pulmonary artery (n = 674), the coefficient of linear correlation (Rlin) between G and SH was 0.982, and the regression line of G was 1.101 x SH-0.072 l/min. When the right atrium was the site of indicator injection (n = 88), Rlin was 0.982, and the regression line nearly agreed with the line of unity. In the artificial circulation model with the single circulation arrangement (n = 81), Rlin between SH and known flow (Q) was 0.959, and between G and Q it was 0.951; both methods over-estimated the flows, with the over-estimation being greater in the case of G. In the artificial model with recirculation effect (n = 31), Rlin between SH and Q was 0.953, and between G and Q 0.942, G again over-estimating the real flow essentially more than SH. Thus, in most cases the Stewart-Hamilton principle seems to be the more reliable way of analysis of dye dilution curves also in modern computer practice.