Requirements and implementation of a flexible kinetic modeling tool

Abstract

Kinetic (or compartment) modeling is a highly versatile tool for the analysis of experiments within living systems. In PET, it is essential for developing tracers, for assessing tracer behavior and for extracting quantitative information about the target process. However, tools to support the modeling tasks involved are not easily available.

Methods: This article presents a requirements analysis for kinetic modeling in PET. The interactive kinetic modeling tool KMZ implements many of these features. It facilitates model development by a set of predefined models and by the ease of introducing new models. Monte Carlo studies allow assessing parameter identifiability. The responses in the different compartments as well as the expected time-activity curve can be simulated for specific model configurations. For measured time-activity curves, model optimization can be performed by the Powell or the Marquardt algorithm. Both support weighted nonlinear least-squares fitting and allow optional constraints of parameter ranges. To further improve parameter estimation, the fitting of several regional time-activity curves can be coupled, resulting in lower standard errors for parameters common among regions. It is possible to highly automate the evaluation of study series and to forward the results into statistical analysis tools.

Results: The KMZ tool has proven highly suitable in evaluating data from different types of studies, and the intuitive user interface enables medical doctors to successfully perform routine evaluations after a short training period.

Conclusion: A portable kinetic modeling tool with the described features would provide easy access to model development and may help consolidate kinetic modeling in clinical settings for well-defined applications.