Research ArticleImaging

A Monte Carlo Study of the Dependence of Early Frame Sampling on Uncertainty and Bias in Pharmacokinetic Parameters from Dynamic PET

Ida Häggström, Jan Axelsson, Charles Ross Schmidtlein, Mikael Karlsson, Anders Garpebring, Lennart Johansson, Jens Sörensen and Anne Larsson
Journal of Nuclear Medicine Technology March 2015, 43 (1) 53-60; DOI: https://doi.org/10.2967/jnmt.114.141754

Article Figures & Data

Figures

  • FIGURE 1.
    FIGURE 1.

    2-tissue-compartment model consisting of compartments of arterial blood plasma (Cp), free-plus-nonspecific (nondisplaceable) and specifically bound tracer in tissue (CF + NS and CS), 4 rate constants (K1, k2, k3, k4), and tissue blood fraction (Va). CPET is the apparent concentration in a PET image VOI or voxel.

  • FIGURE 2.
    FIGURE 2.

    Voxelized BrainWeb head phantom, with inserted spheric blood and tumor regions, labeled with their diameter in mm. All tissues were attributed realistic TACs, and the blood region was assigned the input function (Cp) and all 14 tumor regions the same TTAC (CPET).

  • FIGURE 3.
    FIGURE 3.

    Common input function Cp and CPET for the two 18F-FLT parameter sets, showing the activity concentration for the blood and tumor regions, respectively.

  • FIGURE 4.
    FIGURE 4.

    Example images from the FLT1 study, for the different early frame-sampling schemes plus true phantom image. The frame around 120 s is presented, and the early frame durations are indicated in the columns of figure.

  • FIGURE 5.
    FIGURE 5.

    Representative example from first 20 min of FLT1 study, showing OSEM. TTACs and corresponding NLS fits for 6 different early frame-sampling schemes are shown.

  • FIGURE 6.
    FIGURE 6.

    Relative error (bias) (A) and uncertainty (SD) (B) in all parameter estimates for 6 different early frame-sampling schemes. Error bars in bias represent SE. y-axes are scaled differently for better visibility.

  • FIGURE 7.
    FIGURE 7.

    Relative bias for all model parameters for the 6 different early frame-sampling schemes, using the resampled theoretic (noiseless) input function and TTAC for the NLS fit.

Tables

  • TABLE 1

    2 Sets of Simulated 18F-FLT Parameter Values: FLT1 (12) and FLT2 (13)

    Parameter setK1k2k3k4VaKi*
    FLT10.0710.0910.0470.0180.0860.024
    FLT20.1110.1310.0170.0120.1220.013

    • ↵* Values calculated by Equation 1.

  • TABLE 2

    Total Number of kcounts in the Early Frames (First 120 Seconds of Acquisition) for the Different Sampling Schemes

    Parameter set1 s2 s4 s6 s10 s15 s
    FLT10–440–870–1733–258125–429282–642
    FLT20–570–1130–2264–338162–564368–843
Back to top

In this issue

Print
Download PDF
Article Alerts
Email Article
Citation Tools
Bookmark this article

Jump to section

Keywords