Alternative approach to estimate lumped constant in the deoxyglucose model: simulation and validation.

An alternative method of estimating the lumped constant (LC) in the deoxyglucose model was developed. The LC was estimated using data obtained during the first 10 min after injection of the tracer by a nonlinear least-squares (NLSQ) method. The method does not require a constant plasma concentration. This approach was evaluated in a computer simulation by adding different levels of noise and considering various input functions. Errors in the estimated LC in this and Sokoloff et al.'s conventional method were compared. We found that the approach proposed here results in more reliable estimates of LC. The study in completed in a shorter experimental period, and any shape of the input function can be used. The new technique was then applied to measure whole brain LC and rate constants in cat brain for 2-[18F]fluoro-2-deoxy-D-glucose (2-[18F]FDG). Measured mean value (+/-s.e.m.) for the whole brain LC = 0.443 +/- 0.012 (N = 7), for the whole brain k2* = 0.124 +/- 0.009 and k3* = 0.024 +/- 0.001 (N = 7).