Effects of nonideal input functions on PET measurements of pulmonary blood flow

Regional pulmonary blood flow (rPBF) can be measured with an intravenous infusion of 15O-labeled water and positron emission tomography (PET). The current method depends on two assumptions related to the input of activity to the lung during the scan: 1) the pulmonary arterial tracer input is constant (i.e., a "step function" in shape), and 2) the scan begins at the instant of arrival of the step function. To determine the effect that departures from these assumptions might have on the measurement of rPBF, we performed a series of mathematical simulations for three different input functions: 1) a step function that arrived either 1 or 2 s before or after scan start; 2) a dispersed input function, with activity rising during the scan period; and 3) a combination of these two errors. Calculated values, based on the standard assumptions, were compared against the "known" values used in generating the simulated data. The results show that timing errors associated with starting the scan late cause an overestimation of rPBF, whereas timing errors due to low regional flow or departures from the assumed step input function both cause an underestimation of true rPBF. Thus, in actual practice, the combined errors probably partially offset one another. Except for states of truly high rPBF and low lung density, the errors remain less than 15% of the true value. We conclude that PET measurements of rPBF are not highly sensitive to these presumably common departures from the assumed pulmonary arterial input function to lung regions of interest.