Noninvasive automatic estimation of input function for dynamic mouse 18F-FDG microPET studies

Objectives We present and validate a noninvasive method to estimate the input function (IF) from dynamic 18F-FDG microPET mouse studies.

Methods The image-derived left ventricle (LV) and myocardium time activity curves (TACs) were firstly extracted using constraint nonnegative matrix factorization method automatically. Since the IF equals the ideal TAC in LV, we modeled the ideal TAC in LV with a 7-parameters mathematic equation. The model predicted output of LV and myocardium TACs are more accurately expressed as a linear combination of ideal TAC in LV and myocardium due to the partial volume and spillover effects. Since it has been reported that the liver TAC at late time can be used as an approximation of the blood TAC with proper linear correction, we utilized the information of liver TAC at late time as a Bayesian constraint and incorporated it into the cost function. We also introduced two weighting factors associated with the residuals between the model predicted LV/myocardium TACs and the measured image-derived LV/myocardium TACs into the cost function. Then model parameters and weighting factors were simultaneously estimated by use of an extended least-squares method. After the IF was obtained, it was delayed and dispersed to compare with the golden standard blood samples.

Results Twenty studies from the Mouse Quantitation Program database, which were shared by UCLA, were used to verify our method. The differences of the area under the curves between the estimated IF and blood samples was 8.4% ± 11.8%, and the difference of the influx constant Ki in myocardium was 5.7% ± 10.4% (r = 0.934). The experimental results demonstrate the effectiveness of the proposed method.

Conclusions The proposed method is entirely noninvasive and automatic, while correcting for the spillover, partial volume, delay and dispersion effects.

Research Support This paper is supported by the National Basic Research Program of China (973 Program) under Grant 2011CB707700

• Download figure
• Open in new tab
• Download powerpoint

Parameter estimates for 5 studies. K1~k4: rate constants for myocardium. delay~L3 : parameters in IF model equation (unit: MBq/mL)