Novel parameter estimation methods for ¹¹C-acetate dual-input liver model with dynamic PET

Abstract

The successful investigation of ¹¹C-acetate in positron emission tomography (PET) imaging for marking hepatocellular carcinoma (HCC) has been validated by both clinical and quantitative modeling studies. In the previous quantitative studies, all the individual model parameters were estimated by the weighted nonlinear least squares (NLS) algorithm. However, five parameters need to be estimated simultaneously, therefore, the computational time-complexity is high and some estimates are not quite reliable, which limits its application in clinical environment. In addition, liver system modeling with dual-input function is very different from the widespread single-input system modeling. Therefore, most of the currently developed estimation techniques are not applicable. In this paper, two parameter estimation techniques: graphed NLS (GNLS) and graphed dual-input generalized linear least squares (GDGLLS) algorithms were presented for ¹¹C-acetate dual-input liver model. Clinical and simulated data were utilized to test the proposed algorithms by a systematic statistical analysis. Compared to NLS fitting, these two novel methods achieve better estimation reliability and are computationally efficient, and they are extremely powerful for the estimation of the two potential HCC indicators: local hepatic metabolic rate-constant of acetate and relative portal venous contribution to the hepatic blood flow.