In mathesong/kinfitr: Kinetic Modelling of PET Time Activity Curves

kinfitr : PET Kinetic Modelling using R

Overview

kinfitr is a package for PET Kinetic Modelling Using R. The main goal of this package is to equip PET modellers with great flexibility, while simultaneously making it easier to produce, present and share their results in a highly transparent manner using R and its ecosystem of tools for computational reproducibility.

Installation

This package is currently only available on GitHub. It can be downloaded as follows:

remotes::install_github("mathesong/kinfitr")

There is also a docker container if you would prefer not to have to download everything. If you download Docker, you can pull the container and start it with the following commands. The rstudio session can be opened in a browser pointing to http://localhost:8787.

docker pull mathesong/kinfitr_docker
docker run -e PASSWORD=XYZ --name rstudio -p 8787:8787 mathesong/kinfitr_docker

Alternatively, if you want to mess around with the kinfitr package and its vignettes in order to get a feel for it before installing it, click the "launch binder" button above to open the vignettes in a binder instance.

Citation

At present, there are two preprints out about kinfitr. Please cite at least one of these if you use kinfitr in your study.

An introduction to the package:

Matheson, G. J. (2019). Kinfitr: Reproducible PET Pharmacokinetic Modelling in R. bioRxiv: 755751. https://doi.org/10.1101/755751

A validation study compared against commercial software:

Tjerkaski, J., Cervenka, S., Farde, L., & Matheson, G. J. (2020). Kinfitr – an open source tool for reproducible PET modelling: Validation and evaluation of test-retest reliability. bioRxiv: 2020.02.20.957738. https://doi.org/10.1101/2020.02.20.957738

Background and Usage

I've written up a series of blog posts describing some PET modelling theory and demonstrating how the package can be used. These will eventually be included as vignettes in the package. You can launch them in a cloud instance to play around with them yourself by clicking the "launch binder" button above.

Implemented Models

Reference Region Models

Simplified Reference Tissue Model (SRTM) (Lammertsma & Hume, 1996)
Non-Invasive Logan Plot (Logan et al., 1996)
Non-Invasive Multilinear Logan Plot (Turkheimer et al., 2003)
Ichise Multilinear Reference Tissue Model 1 (MRTM1) (Ichise et al., 2003)
Ichise Multilinear Reference Tissue Model 2 (MRTM2) (Ichise et al., 2003)
Patlak Reference Tissue Model (Patlak & Blasberg, 1985)

Models Requiring Arterial Input

One-Tissue Compartment Model
Two-Tissue Compartment Model
Logan Plot (Logan et al., 1990)
Multilinear Logan Plot (Turkheimer et al., 2003)
Ichise Multilinear Analysis 1 (MA1) (Ichise et al., 2002)
Ichise Multilinear Analysis 2 (MA2) (Ichise et al., 2002)
Patlak Plot (Patlak et al., 1983)

Other Models

Simultaneous Estimation of Non-Displaceable Binding (SIME) (Ogden et al., 2015)

To-Do

General

Vignettes

Additions

Improvements

Tidy up
- Documentation still quite rough and several inconsistencies
- T-star finders have lots of code duplication: should be more generic
Revise vignette
Update plyr functions to dplyr functions
Get the package CRAN-ready