QModeling: a Multiplatform, Easy-to-Use and Open-Source Toolbox for PET Kinetic Analysis

Kinetic modeling is at the basis of most quantification methods for dynamic PET data. Specific software is required for it, and a free and easy-to-use kinetic analysis toolbox can facilitate routine work for clinical research. The relevance of kinetic modeling for neuroimaging encourages its incorporation into image processing pipelines like those of SPM, also providing preprocessing flexibility to match the needs of users. The aim of this work was to develop such a toolbox: QModeling. It implements four widely-used reference-region models: Simplified Reference Tissue Model (SRTM), Simplified Reference Tissue Model 2 (SRTM2), Patlak Reference and Logan Reference. A preliminary validation was also performed: The obtained parameters were compared with the gold standard provided by PMOD, the most commonly-used software in this field. Execution speed was also compared, for time-activity curve (TAC) estimation, model fitting and image generation. QModeling has a simple interface, which guides the user through the analysis: Loading data, obtaining TACs, preprocessing the model for pre-evaluation, generating parametric images and visualizing them. Relative differences between QModeling and PMOD in the parameter values are almost always below 10−8. The SRTM2 algorithm yields relative differences from 10−3 to 10−5 when k2′\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ {k}_2^{\prime } $$\end{document} is not fixed, since different, validated methods are used to fit this parameter. The new toolbox works efficiently, with execution times of the same order as those of PMOD. Therefore, QModeling allows applying reference-region models with reliable results in efficient computation times. It is free, flexible, multiplatform, easy-to-use and open-source, and it can be easily expanded with new models.