Previous Topic

Next Topic

Book Contents

Book Index

Processing Overview

Data processing of studies with PET or SPECT tracers typically consists of the following parts:

1. Data Loading

  • Create a new data set.
  • Load the plasma curve (total activity concentration of tracer in the plasma fraction of the blood samples).
  • Load the parent fraction curve (fraction of unchanged tracer in plasma, optional).
  • Load the whole blood curve (total activity concentration of tracer in the blood samples, optional).
  • Load the average time-activity curves of one or multiple tissue regions.

2. Input Curve Configuration

  • Define interpolation functions (called models) for the different types of blood data and fit them to the data.
  • The input curve is obtained as the multiplication of the plasma model with the parent fraction model.

3. Kinetic Model Configuration

  • Select a region with a "typical" time-activity curve (TAC) from the Region list.
  • Select a simple model from the Model list.
  • Enable the parameters to be fitted by checking their boxes; unchecked parameters remain at the initial values entered.
  • Select a weighting scheme of the residuals; variable or constant (default) weighting is available.

4. Kinetic Model Fitting

  • Fit the model parameters by activating the Fit current region button.
  • Consult the residuals to check whether the model is adequate; there should ideally be no bias in the residuals, just random noise.
  • If the model is fine, configure Copy the model to all regions to Model & Par, activate the button to establish then save initial model configuration for all TACs, then activate Fit all regions.
  • If the model is not yet fine, test more complex models.

5. Kinetic Model Comparison

  • Switch between compartment models of different complexity and fit. The parameters are either maintained for each model type, or converted, according to the Model conversion setting in the Menu.
  • Check the residuals for judging model adequacy.
  • Check the different criteria on the Details tab (Schwartz Criterion SC, Akaike Information Criterion AIC, Model Selection Criterion MSC) to decide whether a more complex model is supported by the data.
  • Check for parameter identifiability. As an indicator of the parameter standard error coefficients of variation (%COV) are returned from the fit. They should remain limited for all relevant parameters. Additionally, Monte Carlo simulations can be performed to obtain statistics of the parameter estimates.
  • If justified by physiology, try to improve the stability of parameter estimation by enforcing common parameters among regions in a coupled fitting procedure.
  • Compare the outcome of compartment models with that of other models, such as reference models or graphical plots.

6. Saving

  • Save all model information together with the data in a composite text .km file which allows restoring a session.
  • Save a summary of all regional parameters in an EXCEL-ready text file .kinPar.

7. Batch Processing

Lengthy calculations such as coupled fits with many regional TACs or Monte Carlo simulations can be run in a batch job. The results are saved both as .km files and in an EXCEL-ready text file.