Relative Strengths of Three Linearizations of Receptor Availability: Saturation, Inhibition, and Occupancy Plots.

ABSTRACT

We derived three widely used linearizations from the definition of receptor availability in molecular imaging with positron emission tomography (PET). The purpose of the present research was to determine the convergence of the results of the 3 methods in terms of 3 parameters-occupancy (s), distribution volume of the nondisplaceable reference binding compartment (VND), and nondisplaceable reference binding potential (BPND) of the radioligand-in the absence of a gold standard. We tested 104 cases culled from the literature and calculated the goodness of fit of the least-squares and Deming II methods of linear regression when applied to the determination of s, VND, and BPND using the goodness-of-fit parameters R2, coefficient of variation (root-mean-square error [RMSE]), and the infinity norm (‖X‖∞) with both regression methods. We observed superior convergence among the values of s, VND, and BPND for the inhibition and occupancy plots. The inhibition plot emerged as the plot with a slightly higher degree of convergence (based on R2, RMSE, and ‖X‖∞ value). With two regression methods (the least-squares method [LSM] and the Deming II [DM] method), the estimated values of s, VND, and BPND generally converged. The inhibition and occupancy plots yielded the best fits to the data, according to the goodness-of-fit parameters, due primarily to absence of commingling of the dependent and independent variables tested with the saturation (original Lassen) plot. In the presence of noise, the inhibition and occupancy plots yielded higher convergences.