CHAPTER 55 – Redundant-Weighted Integration Method for FDG Parametric Image Creation Using Linear Least Square Estimation

Weighted integration is a practical method that may be used to create parametric images, because its calculation time is much faster than an ordinary nonlinear parameter estimation procedure. In weighted integration methods, the number of weight functions equals the number of parameters to be estimated. In this chapter, more weight functions are applied, which means redundant weight functions are used, and parameters are estimated using a linear least squares procedure. A set of simulations was done for the fluorodeoxyglucose 3K model with various types of weight function, various numbers of weight functions, and various noise levels in the tissue time–activity curve. In the case of Chebyschev or Gaussian functions, the standard deviation (SD) of estimated rate constants decreased with an increase in the number of weight functions. One can conclude that redundant weight functions are effective to reduce noise interference in the tissue time–activity curve. The optimal number of weight functions is between 5 and 10, and the Gaussian function is the best of these three types of weight functions.